Abstract: Some implementations are directed to generating a personal database entry for a user based on free-form natural language input formulated by the user via one or more user interface input devices of a computing device of the user. The generated personal database entry may include one or more terms of the natural language input and descriptive metadata determined based on one or more terms of the natural language input and/or based on contextual features associated with receiving the natural language input. Some implementations are directed to generating, based on one or more personal database entries of a user, output that is responsive to further free-form natural language input of the user. For example, one or more entries that are responsive to further natural language input of the user can be identified based on matching content of those entries to one or more search parameters determined based on the further input.

Abstract: A technique for arranging nodes on a landscape based on a viewpoint desired by a user is provided. One aspect of the present disclosure relates to a node processing apparatus for synthesizing and extracting feature quantities that meet the needs of a user analysis from a plurality of types of feature quantities assigned for each node of a node set, and the node processing apparatus includes a receiving unit configured to receive, from a user, a designation related to an arrangement of nodes selected from the node set on an analysis axis assumed by the user, and a node processing unit configured to synthesize and extract feature quantities based on the arrangement of the received designation.

Abstract: A method of identifying and enabling a data marketplace system transaction may comprise receiving an enterprise user object map defining a dataset field name and storage location for a dataset to be migrated pursuant to a user data integration process modeled by a user with a plurality of visual elements, identifying a data marketplace domain classification for the dataset to be migrated based on the enterprise user object map, querying a data marketplace application programming interface (API) to determine that it provides a marketplace for datasets meeting the data marketplace domain classification, generating a dataset sale integration process for transmitting the dataset of the data marketplace domain classification selected by the user for sale from a storage location defined by the user data integration process to a location defined by the data marketplace API, and automatically transmitting the dataset of the data marketplace domain classification to a purchaser.

Abstract: A system for re-ranking retrieved matching images. The system comprises a processor configured to extract at least one QI global feature from query image by global feature extractor module, and extract plurality of QI local features for the query image by local feature extractor module, and then search and retrieve top-k reference images from plurality of reference images based on at least one QI global feature and at least one KRI global feature associated with top-k reference images by filtering module. The processor is configured to perform matching of the plurality of QI local features with the plurality of KRI local features associated with the top-k reference images by fine tuning module and generate matching distance by distance fusion module, wherein the top-k reference images are re-ranked based on the matching distance and generate confidence score for the top-n reference images by confidence score generation module.

Abstract: Provided are a map data processing method and apparatus, and a storage medium, which relate to the field of data processing technology and, in particular, to artificial intelligence technology, for example, computer vision, map technology, and intelligent transportation. A specific implementation solution includes: processing landform coverage data according to a type of surface species coverage and a data processing rule associated with the type of surface species coverage to obtain a surface species coverage effect map; and generating a landform map according to the surface species coverage effect map and a reference map.

Abstract: A method and a computer system for determining the relevance of a text. The method includes: determining the similarity of the text to texts in a portfolio, where the text is compared with each of the texts in the portfolio as part of an individual comparison while determining a similarity value which indicates the similarity between the two texts in each case; assigning the similarity value to one of the two texts in the respectively performed individual comparison which was published at an earlier time or was first captured by a capture system at an earlier time; calculating a relevance value from the similarity values which were determined in the individual comparisons and were assigned to the text, and storing the calculated relevance value and/or transmitting the calculated relevance value to a communication end system via a computer network.

Abstract: An autonomous vehicle includes a sensor detecting an object in a vicinity of the autonomous vehicle, a processor coupled to the sensor to receive sensor data from the sensor, wherein the processor is configured to process the sensor data to recognize that the object is a pedestrian or cyclist. The processor cooperates with the sensor to track the pedestrian or cyclist. The processor determines if the pedestrian or cyclist satisfies a notification criterion and, upon determining that the pedestrian or cyclist satisfies the notification criterion, the processor outputs an illumination signal. A light projector receives the illumination signal and projects a beam of light onto a portion of a road on which the pedestrian or cyclist is currently situated to visually notify the pedestrian or cyclist that the autonomous vehicle is presently tracking the pedestrian or cyclist.

Abstract: Method(s), apparatus and system(s) are provided for entity type identification and/or disambiguation of entities within a corpus of text the method including: receiving one or more entity results, each entity result comprising data representative of an identified entity and a location of the identified entity within the corpus of text; identifying an entity type for each entity of the received entity results by inputting text associated with the location of said each entity in the corpus of text to a trained entity type (ET) model configured for predicting or extracting an entity type of said each entity from the corpus of text; and outputting data representative of the identified entity type of each entity in the received entity results.

Abstract: An example operation may include one or more of connecting, by a sparse peer, to a blockchain network of a plurality of peers controlled by orderers, selecting, by the sparse peer, a subset of data to be replicated, specifying, by the sparse peer, data selection filters based on the selected subset of the data, providing, by the sparse peer, the selection filters to the plurality of the peers, receiving, by the sparse peer, blocks matching the selection filters over a gossip protocol from the plurality of the peers, constructing, by the sparse peer, a local ledger of the sparse peer from the received blocks, and bootstrapping the sparse peer based on the local ledger.

Abstract: The present disclosure is directed towards systems and methods for detecting deviations between documents and portions thereof, extracting information from text and detecting deviations between obligations. Deviations are detected by splitting sentences in two documents apart, matching sentences from the different documents and detecting deviations between the matched sentences.

Abstract: A method for providing one or more random sample documents from a corpus of documents using a search engine is provided. The providing of each of the random sample documents comprises selecting randomly a time window from a set of time windows. A search query is sent to the search engine defining a search for documents of the corpus with time-stamps within the time window defined by the randomly selected time window. In response to the sending of the search query, a search result is receiving from the search engine. The search result comprises a set of the documents of the corpus with time-stamps within the time window. One of the documents comprised by the received set of documents is then selected randomly.

Abstract: The disclosure relates to method and system recommending tuning of parameters to generate a data analytics model. The method includes identifying at a data pre-processing stage a pre-processing subset from an associated set of predefined pre-processing methods for a predefined objective. The method includes identifying at a feature selection stage a feature subset from an associated set of predefined feature selection methods for the predefined objective. The method includes identifying at a model training stage a training subset from an associated set of predefined model training methods for the predefined objective. The method further includes generating a plurality of data analytics tuples and selecting a data analytics tuple from the plurality of data analytics tuples. An output result of the data analytics tuple includes highest ranked results for the predefined objective.

Abstract: A system can include one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform operations comprising: receiving, at a search engine, a user search query submitted by a user; generating, using a recall personalization model, a simulated query that supplements the user search query with a feature vector reflecting personalization preferences of the user; and generating, using the search engine, search results for the user search query based, at least in part, on the simulated query that accounts for the personalization preferences of the user. Other embodiments are disclosed herein.

Abstract: A system is provided for reranking. The system comprises a user device and one or more servers. The system is configured to receive a plurality of candidate lists, rerank the plurality of candidate lists based on page-level information and a format of a recommendation page, generate recommendation results based on the reranked lists, and send the recommendation results to the user device. Each candidate list comprises a plurality of candidate items. The page-level information comprises interactions between the candidate items in each candidate list and between different candidate lists among the plurality of candidate lists. The reranking comprises using the format of the recommendation page to determine pairwise item influences between candidate item pairs among the candidate items in the candidate lists. The user device is configured to display the recommendation page with the recommendation results from the one or more servers.

Abstract: A non-transitory computer-readable medium includes computer-executable instructions that cause a processing unit of a user device to receive a multi-link uniform resource locator (URL), wherein the multi-link URL is associated with a plurality of routing link URLs configured to route the user device to a plurality of corresponding application pages. The executable instructions cause the processing unit to receive user selection of the multi-link URL and rank the routing link URLs based on application installation data stored on the user device, wherein the application installation data indicates whether applications for the routing link URLs are installed on the user device. The executable instructions cause the processing unit to select a final routing link URL based on the ranking of the routing link URLs and access an application page associated with the final routing link URL.

Abstract: The present disclosure relates to systems and methods for parsing unstructured data with neural networks. In one implementation, a system for parsing unstructured data may include at least one processor and at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to: receive unstructured data; apply a classifier to the unstructured data to identify a type of the unstructured data; based on the identification, select a corresponding neural network; apply the selected neural network to the unstructured data to obtain structured data; and output the structured data.

Abstract: Described is a system and method for enabling visual search for information. With each selection of an object included in an image, additional images that include visually similar objects are determined and presented to the user.

Abstract: A noise management system analyzes messages from a sender to a plurality of recipients. The noise management system prepares a message frequency of a quantity of messages sent to a unique recipient over a period of time. The noise management system sends the message frequency with a noise rank to the client device for the recipient to review.

Abstract: As provided herein, a domain model, corresponding to a domain of an image, may be merged with a pre-trained fundamental model to generate a trained fundamental model. The trained fundamental model may comprise a feature description of the image converted into a binary code. Responsive to a user submitting a search query, a coarse image search may be performed, using a search query binary code derived from the search query, to identify a candidate group, comprising one or more images, having binary codes corresponding to the search query binary code. A fine image search may be performed on the candidate group utilizing a search query feature description derived from the search query. The fine image search may be used to rank images within the candidate group based upon a similarity between the search query feature description and feature descriptions of the one or more images within the candidate group.

Abstract: Metric time series (MTS) data objects stored within in-memory storage are marked as inactive in response to determining that no MTS data has been received for the MTS objects within a first predetermined time period. In response to determining that an MTS object has been inactive for longer than a second predetermined time period, the MTS data object is migrated from in-memory storage to on-disk storage. Queries directed to MTS objects are first run against MTS object data stored within in-memory storage, and then against MTS object data stored within on-disk storage. In this way, an amount of in-memory storage needed to store MTS objects may be minimized, while optimizing search performance.

Abstract: The present disclosure provides systems and methods for content quasi-personalization or anonymized content retrieval via aggregated browsing history of a large plurality of devices, such as millions or billions of devices. A sparse matrix may be constructed from the aggregated browsing history, and dimensionally reduced, reducing entropy and providing anonymity for individual devices. Relevant content may be selected via quasi-personalized clusters representing similar browsing histories, without exposing individual device details to content providers.

Abstract: A system, server device, and method are provided for managing document metadata. The method includes storing metadata in a graph separately from documents with which the metadata is associated and connecting metadata in the graph according to a hierarchy, with each node of the graph being classified based on metadata content. The method also includes assigning properties to edges in the graph connecting the nodes; receiving, via a communications module, an instruction to update the metadata; and dynamically updating at least one edge or node in the graph according to the instruction without requiring access to an associated document. The method also includes receiving, via the communications module, an inquiry related to a document; and responding to the inquiry, via the communications module, after accessing metadata associated with the document by navigating the graph.

Abstract: A computing device of a system is deployed within a communication network for processing environmental, social, and governance (ESG) reports. The computing device is configured to obtain an ESG report of an entity, receive a question in natural language regarding information contained in or inferable from the ESG report, convert the ESG report into page-wise pieces of text, determine an answer to the question based at least upon the page-wise pieces of text of the ESG report, determine a location of a snippet related to the answer in the page-wise pieces of text of the ESG report, generate a bounding box through which the snippet is highlighted in a page containing the snippet, crop the page based on the bounding box, and display a cropped image of the page.

Abstract: A search query and a concept comprising a feature are received from a user. Positive and negative item sets are identified, where items of the positive item set comprise the feature and items of the negative item set do not comprise the feature. A hyperplane located between the positive and negative item sets is generated using a machine learning model. A concept activation vector (CAV) orthogonal to the hyperplane is determined. The CAV is used to produce a modified search query vector. Based on comparing the modified search query vector with item vectors, a second set of search results is provided.

Abstract: Search-result explanation systems, methods, and computer-program products receive a user search query, expand the search query into a plurality of sub-queries, perform a database search using the expanded user search query, and determine which sub-queries of the plurality of sub-queries matched with a particular search result. Results from the database search are re-indexed in an index generated on-the-fly and in-memory, within which the results are searched using the sub-queries to determine matching fields and match types. A score is determined based on the type of match(es) with a particular search result based on one or more predefined weights and normalized using a denominator comprising a fictitious, on-the-fly record configured to receive a perfect score according to the received user search query. A user interface showing ranked results and explanations for the ranking, including a score for the result based on the expanded user search query.

Abstract: Systems and methods are disclosed for processing electronic content, such as text, videos, and images. According to certain embodiments, user interactions with electronic content may be tracked over a plurality of modalities, such as web pages, email, mobile applications, and social media. The tracked user interactions may include copy/paste events, explicit user highlighting, social sharing, and user voting. Key passages of electronic content may be identified based on the tracked user interactions and ranked against one another. Ranking of passages may be based, for example, on a raw or normalized score for the identified key passages. Alternatively, the ranking of a passage may be based on a ratio of user interactions with the passage to total views of the electronic text containing the passage. One or more of the identified key passages (e.g., the highest ranked passages) may be published to one or more applications.

Abstract: Techniques described herein elect how data is deduplicated in a storage system. A similarity hash signature for a data unit is calculated. A digest table is searched for a similarity hash signature within a predetermined distance of the similarity hash signature for the data unit. Based on the search, either a similarity hash signature or a strong hash signature of the data unit is added to the digest table.

Abstract: Implementations are described herein for recommending actions based on entity or entity type. In various implementations, a partial free-form natural language input may be received from a user at an input component of a computing device. The partial free-form natural language input may identify an entity without identifying a responsive action and may be directed by the user to an automated assistant that operates at least in part on the computing device. The partial free-form natural language input may be analyzed to identify the entity. Based on the identified entity, a plurality or superset of candidate responsive actions may be identified, filtered, and/or ranked based on one or more signals. The automated assistant may then provide output that recommends one or more of the candidate responsive actions based on the ranking and/or filtering.

Abstract: A method, apparatus and device used to search for content, and a computer-readable storage medium, which relate to the field of data processing. In the method, in response to receiving a search request for a target search term, obtain multiple historical search records related to multiple historical search requests, each historical search record comprising the historical search term targeted by the corresponding historical search request. Then, determine a first historical search record matching the target search term from the multiple historical search records. Furthermore, based on the relationship between multiple historical search records, determine from the multiple historical search records a second historical search record associated with the first historical search record. The method further comprises determining an extended result for the target search term on the basis of the search result corresponding to the second historical search record.

Abstract: Systems and methods are provided for determining an environmental impact of one or more transactions. An example method generally includes accessing transaction data representative of a plurality of transactions, where each of the transactions involves a user and a merchant, and accessing at least one index indicative of environmental impact of a plurality of merchants. The method also includes generating a graph based on the users and merchants included in the transaction data, whereby the graph is representative of the plurality of transactions, and determining a mapping between ones of the merchants involved in the transaction data and the environmental impact indicated by the at least one index, based on at least one of: the graph, the accessed at least one index, a graph convolution network (GCN), and a graph neural network (GNN). The method then includes publishing the mapping between the ones of the merchants and the environmental impact.

Abstract: A method implements a dataset rank metric for measuring dataset relevance. Metadata is identified for a plurality of datasets. A graph structure is generated in storage. The graph structure includes a multitude of nodes connected by a multitude of edges. Each node of the multitude of nodes representing a respective dataset of a multitude of datasets, the multitude of edges connecting the multitude of nodes according to a data lineage determined from metadata of the multitude of datasets. A composite score is generated for each node of the graph. The computer processor iteratively processes the composite scores for the multitude of nodes of the graph to generate a dataset rank for each dataset. The multitude of datasets is presented in an interface, sorted according to the respective dataset rank of each dataset.

Abstract: Disclosed is a method and system for a single broadband portal where all content may be delivered to users of varied electronic devices. An example embodiment (i) provides a user interface for enabling user selection of media contents per varied electronic devices, each varied electronic device having a respective platform, the user interface indicating one or more different possible media contents; (ii) accepts from a user requests for distribution service to any of the varied electronic devices of the user, the requests specifying a user desired media content; (ii) locates the user desired media content from a plurality of storage modules; (iv) transmits the user desired media content to at least one of the varied electronic devices of the user; and (v) tracks the transmission of the user desired media content for billing purposes.

Abstract: Embodiments of the present invention provide systems, methods, and computer storage media for identifying candidate boundaries for video segments, video segment selection using those boundaries, and text-based video editing of video segments selected via transcript interactions. In an example implementation, boundaries of detected sentences and words are extracted from a transcript, the boundaries are retimed into an adjacent speech gap to a location where voice or audio activity is a minimum, and the resulting boundaries are stored as candidate boundaries for video segments. As such, a transcript interface presents the transcript, interprets input selecting transcript text as an instruction to select a video segment with corresponding boundaries selected from the candidate boundaries, and interprets commands that are traditionally thought of as text-based operations (e.g., cut, copy, paste) as an instruction to perform a corresponding video editing operation using the selected video segment.

Abstract: The present application provides methods, devices and computer readable media for intrinsic popularity evaluation and content compression based thereon. In an embodiment, there is provided a method of intrinsic popularity evaluation. The method comprises: receiving an image from a social network; and determining an intrinsic popularity score for the image using a deep neural network (DNN) based intrinsic popularity assessment model.

Abstract: Systems and techniques for retrieving video data associated with a selected attribute are described. The systems and techniques include receiving a multimodal input associated with the attribute for querying a catalog of video data to identify video data including the attribute. A first embedding is determined for the input using a first encoder to map the input to a representation space. A second embedding is determined for the video data to map the video data to the representation space. A similarity score is determined between the video data and the input based on a distance between the embeddings. The video data associated with the attribute may be selected based on the similarity score.

Abstract: A computer server system and method are disclosed for personalization and customization of network search results and rankings, such as for Internet searching. A representative server system comprises: a network interface to receive a query from a user and transmit return queries and search results; a data storage device having a first, lexical database having one or more compilations and templates; and one or more processors configured to access the first database and search a selected compilation using the query to generate initial search results; to comparatively score each selected parsed phrase of the initial search results, for each classification of a selected template and a selected compilation, and to output initial and final search results arranged according to the classifications and the predetermined order of the template. A representative embodiment may also include use of a second, semantic database having multi-dimensional vectors corresponding to parsed phrases, paragraphs, or clauses.

Abstract: Another aspect of the invention provides a system for identifying causes of loss from insurance claims data comprising a plurality of unstructured or semi-structured insurance claims data. The system includes: a processor; and computer-readable memory containing instructions to: implement an interface programmed to receive insurance claims data comprising one or more insurance claims; store the insurance claims data in the computer-readable memory; and invoke execution of the method as described herein on the processor.

Abstract: The disclosed embodiments relate to systems and methods for generating an optimal solution for determining a value for one or more base data objects. A plurality of solutions include one or more composite data objects transacted by a transaction system processor. The composite data objects include the one or more base data objects. An optimal solution is generated by using data indicative of a level of activity and a number of sources for each of the plurality of composite data objects.

Abstract: This application relates to ransomware detection and data pruning management. Ransomware typically involves an I/O heavy process of encrypting data files and/or deleting or renaming the original files. Thus, ransomware attacks may be detected by analyzing the I/O activity in a given file system. In some embodiments, a software module running on a client machine manages copying, archiving, migrating, and/or replicating of primary data and restoring and/or pruning secondary data (e.g., backup copies of the primary data). When a potential ransomware attack is detected, the software module is immediately stopped so that the software module does not prune any data that may need to be restored. Upon receiving user input that indicates that the client machine is not under a ransomware attack, the software module is allowed to resume its operations, including pruning of the secondary data.

Abstract: Aspects described herein may allow for generating a filter list to facilitate users to refine their online search. Data items displayed on a web page may be determined based on the user's input data, for example, based on a selection command selecting the data items. Each data item corresponds to a feature of a product. Filters corresponding to the data items may be determined and the weight of each filter may be updated. A filter list may be generated and displayed to facilitate the user to refine his search for a class of products. The filters on the filter list may be automatically ordered based on the weight of each filter. In this way, a user may be able to find filters that are associated with features that he cares about the most and therefore use the filter list to obtain more satisfying results.

Abstract: A system and method for evaluating the performance and usage of a cognitive computing tool which answers questions from users. A log file for these interactions includes the questions, the answers and a confidence rating assigned by the tool to each answer. Questions and answers are analyzed to determine validity, accuracy, and categories by subject matter experts or text analytics tools, and the results are added to the log file. Comments and sentiments from users may be analyzed and added to the log file. Additional data about the users, such as identities, demographics, and locations, may be added. Data from the log file may be presented in a dashboard display as metrics, such as trends and comparisons, describing the usage and performance of the cognitive computing tool. Answers may be displayed as they were presented to the users. Selectable filters may be provided to control the data displayed.

Abstract: A computer-implemented method is provided that includes receiving a search query and, responsive to the search query, providing one or more textual comments relevant to the search query. This includes tokenizing the search query and calculating a set of query term frequency metrics. A set of records relevant to the search query is then selected, from a persistent storage, based on determined similarities between the query term frequency metrics and frequency metrics determined for the records in the persistent storage. Textual comments within the selected records are associated with usefulness metrics. The textual comments relevant to the search query are selected by selecting those textual comments within the selected records that are associated with usefulness metrics that are within a pre-determined range, e.g., an inter-quartile range for a population of usefulness metrics.

Abstract: Examples are directed to systems and methods that cause a first user interface having an input field that receives a search query to be displayed. A search query having an attribute of a first entity is received at the input field. Data records are searched for a second entity using the attribute when the search query is received. A second user interface is displayed that has a data record for the second entity. The data record lists the attribute and an additional attribute of the second entity. The second user interface has a drill down user interface that lists the attribute and the additional attribute. The attribute and the additional attribute are selectable to reconfigure the second user interface. The second user interface has a selectable option to create a new data record where the option corresponds to an indication that the first entity is different from the second entity.

Abstract: A user request for a set of images is received via a graphical user interface, the subset of images having an initial order. A subset of neighboring images is selected, wherein the subset of neighboring images is ordered according to a first sequence. A first concordance value characterizing visual detectability of one or more illicit content features across images in the subset of neighboring images ordered according to the first sequence is computed. A second concordance value characterizing visual detectability of one or more illicit content features across images in the subset of neighboring images ordered according to a second sequence is computed. Responsive to a difference between the second concordance value and the first concordance value satisfying a predetermined condition, the subset of neighboring images is rearranged according to the second sequence.

Abstract: Visually guided machine-learning language model and embedding techniques are described that overcome the challenges of conventional techniques in a variety of ways. In one example, a model is trained to support a visually guided machine-learning embedding space that supports visual intuition as to “what” is represented by text. The visually guided language embedding space supported by the model, once trained, may then be used to support visual intuition as part of a variety of functionality. In one such example, the visually guided language embedding space as implemented by the model may be leveraged as part of a multi-modal differential search to support search of digital images and other digital content with real-time focus adaptation which overcomes the challenges of conventional techniques.

Abstract: A categorization analysis system is provided. The categorization analysis system includes one or more hardware processors, a memory including a first plurality of listings categorized in a first target category, and a categorization analysis engine executing on the one or more hardware processors. The categorization analysis engine is configured to determine a label for each listing including performing a search on title, select a set of training listings based on the determined labels, train a first model using the set of training listings and the determined labels, the first model being a classification model configured to classify categorization of listings, identify a suspect listing categorized in the first target category, apply the suspect listing to the first model, thereby generating a categorization result for the suspect listing, the categorization result indicating miscategorization of the suspect listing, and identify the suspect listing in the memory as miscategorized.

Abstract: Techniques for summarization of search results are provided. A similarity search query comprising at least one search criteria is received. The similarity search query is executed on at least one data source containing a plurality of images associated with metadata responsive to the search criteria. A plurality of search results response to the similarity search query is received from the at least one data source. The plurality of search results is clustered based on the metadata associated with the plurality of search results excluding the similarity criteria. The plurality of search results is summarized based on the results of the clustering. The summarization is displayed in a display view. An interactive user interface is provided to refine the plurality of search results based on the summarization.

Abstract: A computer-implemented method performed at least in part by a graph database configured to store at least one graph dataset. The method includes receiving a graph query configured to be performed against a machine learning model, and communicating the graph query with a machine learning system that is configured to use the machine learning model to obtain model inference results and communicate those model inference results to the graph database application. The graph database provides query results based at least in part on the model inference results to an entity.

Abstract: A collaborative system that securely enables the tracking of changes in an agreement.

Abstract: A plurality of HyperText Markup Language (HTML) strings corresponding to a dataset of document object model (DOM) nodes is tokenized according to a tokenization scheme to produce a dictionary of tokens that occur in the dataset. A condensed dictionary of tokens in produced by removing low-value tokens from the vocabulary of tokens. An information matrix is computed based on the condensed dictionary of tokens, the information matrix being a set of values, a value of the set of values corresponding to a frequency of co-occurrence of a pair of tokens. A library of word vectors is derived from the information matrix. A feature vector of an HTML element is generated based at least in part on the library of word vectors, and a classification for the HTML element is obtained from a machine learning model as a result of inputting the feature vector into the machine learning model.