Automated language model-based training framework for machine learning classifiers

A system comprising:

• one or more processors;
• and
• one or more non-transitory computer readable media storing processor-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
• receiving a data request that includes a search parameter to be applied to a plurality of data elements, wherein the search parameter is associated with an entity identifier that provides a link between two or more data elements;
• generating, via an encoder language model, a search embedding corresponding to the search parameter;
• at a first stage, detecting a first subset of data elements among the plurality of data elements that comprise corresponding embeddings that align with the search embedding in accordance with a first threshold;
• at a second stage, detecting a second subset of data elements among the first subset of data elements that comprise corresponding classification scores that satisfy a second threshold, wherein detecting the second subset of data elements comprises using a machine learned entity classifier model that is trained with a training dataset generated based on an output of the encoder language model, wherein the training dataset comprises:
• (1) a first sample data element (a) labeled with a first pseudo label and (b) comprising a first sample embedding that aligns with the search embedding in accordance with a third threshold that is associated with a greater similarity than the first threshold;
• and
• (2) a second sample data element (a) labeled with a second pseudo label and (b) comprising a second sample embedding that differs from the search embedding in accordance with a fourth threshold that is associated with a lesser similarity than the first threshold;
• and
• in response to the data request, transmitting one or more data packets that comprise one or more data elements of the second subset of data elements.

One or more non-transitory computer-readable media storing processor-executable instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

• receiving a data request that includes a search parameter to be applied to a plurality of data elements, wherein the search parameter is associated with an entity identifier that provides a link between two or more data elements;
• generating, via an encoder language model, a search embedding corresponding to the search parameter;
• at a first stage, detecting a first subset of data elements among the plurality of data elements that comprise corresponding embeddings that align with the search embedding in accordance with a first threshold;
• at a second stage, detecting a second subset of data elements among the first subset of data elements that comprise corresponding classification scores that satisfy a second threshold, wherein detecting the second subset of data elements comprises using a machine learned entity classifier model that is trained with a training dataset generated based on an output of the encoder language model, wherein the training dataset comprises:
• (1) a first sample data element (a) labeled with a first pseudo label and (b) comprising a first sample embedding that aligns with the search embedding in accordance with a third threshold that is associated with a greater similarity than the first threshold;
• and
• (2) a second sample data element (a) labeled with a second pseudo label and (b) comprising a second sample embedding that differs from the search embedding in accordance with a fourth threshold that is associated with a lesser similarity than the first threshold;
• and
• in response to the data request, transmitting one or more data packets that comprise one or more data elements of the second subset of data elements.