Abstract: Systems and methods for generating issue networks are disclosed. In one embodiment, a computer-implemented method of generating an issue network from a document corpus includes searching, using a computer, the document corpus for a set of documents discussing a starting issue, wherein the starting issue is one of a plurality of normalized issues defined by the document corpus. The method further includes determining a set of normalized issues discussed by the set of documents discussing the starting issue, wherein the set of normalized issues also includes the starting issue, and determining instances of co-occurrences of individual normalized issues of the set of normalized issues within individual cases of the set of documents. The method also includes linking individual normalized issues of the set of normalized issues based on their co-occurrences within the set of documents, wherein the linked individual normalized issues at least in part define the issue network.

Abstract: A computer-implemented method of annotating an electronic document may include receiving annotation information corresponding to a first electronic document file and creating annotation metadata that is associated with the annotation information. The method may further include storing the annotation information and associated annotation metadata in an annotation file that is separate from the first electronic document file, and anchoring the annotation information to a target electronic document file at an anchor location corresponding to the annotation metadata. The annotation metadata may be generated by assigning a target offset value to individual neighboring tokens defining an annotation neighborhood, wherein the target offset values correspond to positions of the neighboring tokens with respect to an annotation location within the first electronic document file.

Abstract: A computer-automated system and method identify text in a first “citing” court case, near a “citing instance” (in which a second “cited” court case is cited), that indicates the reason(s) for citing (RFC). The automated method of designating text, taken from a set of citing documents, as reasons for citing (RFC) that are associated with respective citing instances of a cited document, has steps including: obtaining contexts of the citing instances in the respective citing documents (each context including text that includes the citing instance and text that is near the citing instance), analyzing the content of the contexts, and selecting (from the citing instances' context) text that constitutes the RFC, based on the analyzed content of the contexts. A related computer-automated system and method selects content words that are highly related to the reasons a particular document is cited, and gives them weights that indicate their relative relevance.

Abstract: A computer-automated system and method identify text in a first “citing” court case, near a “citing instance” (in which a second “cited” court case is cited), that indicates the reason(s) for citing (RFC). The automated method of designating text, taken from a set of citing documents, as reasons for citing (RFC) that are associated with respective citing instances of a cited document, has steps including: obtaining contexts of the citing instances in the respective citing documents (each context including text that includes the citing instance and text that is near the citing instance), analyzing the content of the contexts, and selecting (from the citing instances' context) text that constitutes the RFC, based on the analyzed content of the contexts. A related computer-automated system and method selects content words that are highly related to the reasons a particular document is cited, and gives them weights that indicate their relative relevance.

Abstract: A computer-automated system and method identify text in a first “citing” court case, near a “citing instance” (in which a second “cited” court case is cited), that indicates the reason(s) for citing (RFC). The automated method of designating text, taken from a set of citing documents, as reasons for citing (RFC) that are associated with respective citing instances of a cited document, has steps including: obtaining contexts of the citing instances in the respective citing documents (each context including text that includes the citing instance and text that is near the citing instance), analyzing the content of the contexts, and selecting (from the citing instances' context) text that constitutes the RFC, based on the analyzed content of the contexts. A related computer-automated system and method selects content words that are highly related to the reasons a particular document is cited, and gives them weights that indicate their relative relevance.

Abstract: A computer-implemented method of gathering large quantities of training data from case law documents (especially suitable for use as input to a learning algorithm that is used in a subsequent process of recognizing and distinguishing fact passages and discussion passages in additional case law documents) has steps of: partitioning text in the documents by headings in the documents, comparing the headings in the documents to fact headings in a fact heading list and to discussion headings in a discussion heading list, filtering from the documents the headings and text that is associated with the headings, and storing (on persistent storage in a manner adapted for input into the learning algorithm) fact training data and discussion training data that are based on the filtered headings and the associated text.

Abstract: A system and method for binary classification of text units such as sentences, paragraphs and documents as either a rule of law (ROL) or not a rule of law (˜ROL). During a training phase of the system and method of the present invention, an initialized knowledge base and labeled or pre-classified sentences are used to build a trained knowledge base. The trained knowledge base contains an equation, a threshold, and a plurality of statistical values called Z values. When inputting text documents for classification, a Z value is generated for each term or token in the input text. The Z values are input to the equation which calculates a score for each sentence. Each calculated score is then compared to the threshold to classify each sentence as either ROL or ˜ROL.

Abstract: An economic, scalable machine learning system and process perform document (concept) classification with high accuracy using large topic schemes, including large hierarchical topic schemes. One or more highly relevant classification topics is suggested for a-given document (concept) to be classified. The invention includes training and concept classification processes. The invention also provides methods that may be used as part of the training and/or concept classification processes, including: a method of scoring the relevance of features in training concepts, a method of ranking concepts based on relevance score, and a method of voting on topics associated with an input concept. In a preferred embodiment, the invention is applied to the legal (case law) domain, classifying legal concepts (rules of law) according to a proprietary legal topic classification scheme (a hierarchical scheme of areas of law).

Abstract: A statistical thesaurus is built dynamically, from the same text collection that is being searched, allowing improved generation of expanded query terms. The thesaurus is dynamic in that thesaurus records are collected, ranked, accessed, and applied dynamically. Thesaurus "records" are actually formed as indexed documents arranged in "collections". The collections are preferably distinguished based on text source (court cases versus news wires versus patents, and so forth). Each record has terms assembled in indexed groups (or segments) which inherently reflect a ranking based on relevance to an initial query. After an initial query is received, the appropriate collection(s) of records may be searched by a conventional search and retrieval engine, the searches inherently returning records ranked by degree of relevance due to the record indexing scheme. A record ranking scheme avoids contamination of relevant records by less relevant records.

Abstract: A phrase recognition method breaks streams of text into text "chunks" and selects certain chunks as "phrases" useful for automated full text searching. The phrase recognition method uses a carefully assembled list of partition elements to partition the text into the chunks, and selects phrases from the chunks according to a small number of frequency based definitions. The method can also incorporate additional processes such as categorization of proper names to enhance phrase recognition. The method selects phrases quickly and efficiently, referring simply to the phrases themselves and the frequency with which they are encountered, rather than relying on complex, time-consuming, resource-consuming grammatical analysis, or on collocation schemes of limited applicability, or on heuristical text analysis of limited reliability or utility.

Abstract: A computer system uses a legal hierarchy annotated with seed citations to generate a control file, and then using the control file, permits legal documents to be classified automatically into the legal hierarchy without the need for manual intervention. Each classification within the legal hierarchy receives a unique numerical classification key which identifies the location of the classification within the legal hierarchy. Each level of the hierarchy also receives a unique hierarchy location key which identifies a hierarchical document through which a user can retrieve a legal document which displays to the user a classification. The control file is an automatically-generated intermediate file which identifies the legal classifications, their classification keys, and the hierarchy location keys to which the classifications map. This automatically generated control file is input to a legal classification generator, along with a document to be classified.