The disclosure relates to AI-based machine-learning and natural language modeling to identify semantic similarities between sets of content having natural language text. For example, a system may generate a relevance classification that indicates whether content such as articles are non-specifically relevant to charities without identifying a particular charity. If the content is non-specifically relevant to charities, the system may apply a natural language model to generate sentence embeddings based on the content and determine a level similarity between the sentence embeddings and a query embedding generated from a charity query. The charity query may itself be generated from a full description of the charity through an encoder-decoder architecture with reinforcement learning.

The disclosure relates to AI-based machine-learning and natural language modeling to identify semantic similarities between sets of content having natural language text. For example, a system may generate a relevance classification that indicates whether content such as articles are non-specifically relevant to charities without identifying a particular charity. If the content is non-specifically relevant to charities, the system may apply a natural language model to generate sentence embeddings based on the content and determine a level similarity between the sentence embeddings and a query embedding generated from a charity query. The charity query may itself be generated from a full description of the charity through an encoder-decoder architecture with reinforcement learning.

A method, computer program, and computer system is provided for identifying a speaker in at text based work. Labeled and unlabeled instances corresponding to one or more speakers are extracted. Pseudo-labels are inferred for the extracted unlabeled instances based on the labeled instances. One or more of the unlabeled instances are labeled based on the inferred pseudo-labels.

A method, computer program, and computer system is provided for identifying a speaker in at text based work. Labeled and unlabeled instances corresponding to one or more speakers are extracted. Pseudo-labels are inferred for the extracted unlabeled instances based on the labeled instances. One or more of the unlabeled instances are labeled based on the inferred pseudo-labels.

A system to quantify aggregate alignment of segmented text with an evaluator population, with a data processing system comprising memory and one or more processors, can segment a first extended text object into one or more evaluation text objects associated with a population reference, identify one or more text frame objects corresponding to the evaluation text objects, the text frame objects being associated with a second extended text object, generate, based on the text frame objects, one or more context identifier objects corresponding to the evaluation text objects, and generate a condensed text object including one or more of the evaluation text objects, the evaluation text objects being positioned in the condensed text object in response to output of a first machine learning model trained with input including at least one of the first extended text objects, the evaluation text objects, the context identifier objects, and the text frame objects.

A system to quantify aggregate alignment of segmented text with an evaluator population, with a data processing system comprising memory and one or more processors, can segment a first extended text object into one or more evaluation text objects associated with a population reference, identify one or more text frame objects corresponding to the evaluation text objects, the text frame objects being associated with a second extended text object, generate, based on the text frame objects, one or more context identifier objects corresponding to the evaluation text objects, and generate a condensed text object including one or more of the evaluation text objects, the evaluation text objects being positioned in the condensed text object in response to output of a first machine learning model trained with input including at least one of the first extended text objects, the evaluation text objects, the context identifier objects, and the text frame objects.

In an example, a first set of text may be received from a client device. A set of content items may be selected from among content items based upon the first set of text and a plurality of sets of content item text associated with the content items. A set of terms may be determined based upon the first set of text and the set of content items. A similarity profile associated with the set of terms may be generated. The similarity profile is indicative of similarity scores associated with similarities between terms of the set of terms. Relevance scores associated with the set of terms may be determined based upon the similarity profile. One or more search terms may be selected from among the set of terms based upon the relevance scores. A search may be performed based upon the one or more search terms.

Provided is a method including obtaining a first data object including a first set of data entries, wherein each data entry of the first set of data entries includes text content associated with a time entry. The method includes generating a first data object score using the text content and the time entries included in the first set of data entries and using scoring parameters, determine that the first data object score satisfies a data object score condition; perform in response to the first data object score satisfying the data object score condition, a condition-specific action associated with the data object score condition.

Provided is a method including obtaining a first data object including a first set of data entries, wherein each data entry of the first set of data entries includes text content associated with a time entry. The method includes generating a first data object score using the text content and the time entries included in the first set of data entries and using scoring parameters, determine that the first data object score satisfies a data object score condition; perform in response to the first data object score satisfying the data object score condition, a condition-specific action associated with the data object score condition.

Disclosed are a method for training speech recognition model, a method and a system for speech recognition. The disclosure relates to field of speech recognition and includes: inputting an audio training sample into the acoustic encoder to represent acoustic features of the audio training sample in an encoded way and determine an acoustic encoded state vector; inputting a preset vocabulary into the language predictor to determine text prediction vector; inputting the text prediction vector into the text mapping layer to obtain a text output probability distribution; calculating a first loss function according to a target text sequence corresponding to the audio training sample and the text output probability distribution; inputting the text prediction vector and the acoustic encoded state vector into the joint network to calculate a second loss function, and performing iterative optimization according to the first loss function and the second loss function.