Abstract: A method of generating keyword-based dialogue summaries is provided. The method includes inputting a transcript of an audio conversation and a keyword into a machine learning model trained based on encodings representing the keyword and the transcript, generating computer-generated text different from and semantically descriptive of the transcript and semantically associated with the keyword, and outputting the computer-generated text in association with a selectable item selectable for inclusion of the computer-generated text in displayed text representing the transcript, the selectable item associated with the keyword.

Abstract: The disclosure herein describes interpreting attention-based decisions of summarization outputs generated by a deep learning model. A decision interpretation model obtains attention values defining connections between input tokens associated with a source text and output tokens for a selected portion of a summary associated with the source text. The input tokens having the highest attention values indicating the strongest connections between the input tokens of the source text and an output token of the summary are selected as primary tokens. A semantic similarity between the primary tokens for each attention head and an output token is calculated. The model selects the primary tokens having the closest semantic similarity with the summary portion. A visual cue is generated on or within a portion of the source text corresponding to the primary tokens. The visual cue identifies dominant words in the source text used to explain the summary portion.

Abstract: A computerized method categorizes questions from a transcript and provides those categorized questions to targets. A transcript associated with a meeting is obtained and a question in the obtained transcript is detected by a question detection model. A question vector embedding of the detected question is generated using a question clustering model and a category of the detected question is determined by the question clustering model using the generated question vector embedding and a plurality of category vector embeddings. A target of the detected question is identified using the determined category and the detected question is provided to the identified target via a question provision interface. Further, the question clustering model is used to generate the category vector embeddings from example questions of category question catalogs, such that the set of categories for which the question clustering model is configured can be efficiently modified.

Abstract: Example solutions for performing key phrase extraction from content items using a large language model (LLM) include: determining a token entropy score for a first token of a content item containing text content by generating and submitting a prompt to the LLM, that includes prefix tokens preceding a first token, receiving a probability distribution from the LLM, and generating a token entropy score for the first token; identifying a candidate phrase that includes one or more tokens, each token having an associated token entropy score; computing a phrase entropy score for the candidate phrase based on the token entropy scores of the one or more tokens; storing the candidate phrase as a key phrase of the content item upon the phrase entropy score exceeding a threshold; and searching a database of content items based on the key phrase, the search returning results including the content item.

Abstract: Example solutions for reducing the likelihood of hallucinations by language models, such as large language models (LLMs) are disclosed. By injecting a sufficient range and quantity of curated factual data into a prompt, the likelihood of a hallucination by an LLM may be reduced. This enables language models to be used in a wider range of settings, in which fabrication of facts is problematic, while reducing the need for a human to carefully check the generated text for accuracy. Examples include: generating a summary of a transcript using a summarization model; extracting topic-specific data from stored data using a scoring model; dynamically generating a language model prompt using the topic-specific data and the summary; and generating an output text using a language model and the language model prompt.

Abstract: Text coherence is classified by receiving a multiword text string into a machine learning model, determining, by the machine learning model, semantic probability data representing a probability that a word of the received multiword text string is semantically correlated to one or more other words in the multiword text string, determining, by the machine learning model, an inferential aggregate perplexity score of the multiword text string, based on the determined semantic probability data, outputting, from the machine learning model, the inferential aggregate perplexity score, and classifying a coherence of the multiword text string based on whether the outputted inferential aggregate perplexity score satisfies a coherence condition, wherein the coherence condition is based on a predefined coherence score.

Abstract: A method of generating keyword-based dialogue summaries is provided. The method includes inputting a transcript of an audio conversation and a keyword into a machine learning model trained based on encodings representing the keyword and the transcript, generating computer-generated text different from and semantically descriptive of the transcript and semantically associated with the keyword, and outputting the computer-generated text in association with a selectable item selectable for inclusion of the computer-generated text in displayed text representing the transcript, the selectable item associated with the keyword.