Topic segmentation of image-derived text
Described herein are systems, methods, and other techniques for segmenting an input text. A set of tokens are extracted from the input text. Token representations are computed for the set of tokens. The token representations are provided to a machine learning model that generates a set of label predictions corresponding to the set of tokens. The machine learning model was previously trained to generate label predictions in response to being provided input token representations. Each of the set of label predictions indicates a position of a particular token of the set of tokens with respect to a particular segment. One or more segments within the input text are determined based on the set of label predictions.
1. A computer-implemented method comprising:
receiving an image of a historical document, the image digitized by optical character recognition of the historical document, the historical document comprising a set of non-narrative units, each non-narrative unit describing one or more natural persons that are different from other non-narrative units and each non-narrative unit concerning the same topic as other non-narrative units;
extracting a set of tokens from the historical document, wherein the set of tokens comprising a first token that represents a particular natural person in a particular non-narrative unit, a second token that represents a relationship associated with the particular natural person;
generating a plurality of embedding vectors representing the set of tokens;
providing the plurality of embedding vectors to a machine learning language model that is trained to divide the set of non-narrative units to individual units; and
displaying a location of the particular non-narrative unit in the image.
2. The computer-implemented method of claim 1 , further comprising:
identifying a set of segments in the image of the historical document; and
identifying the set of non-narrative units based on the set of segments.
3. The computer-implemented method of claim 1 , further comprising:
computing a position vector for each of the set of tokens, wherein the position vector indicates a location of a token with respect to a physical reference point within the image.
4. The computer-implemented method of claim 3 , wherein the position vector of a particular token with respect to a particular segment is one of:
at a beginning of the particular segment;
inside the particular segment; or
outside the particular segment.
5. The computer-implemented method of claim 1 , wherein the historical document is a marriage record that lists couple names as the set of non-narrative units.
6. The computer-implemented method of claim 5 , wherein the marriage record includes a plurality of marriage announcements captured from a newspaper.
7. The computer-implemented method of claim 1 , wherein the machine learning language model includes a bi-directional long short-term memory (LSTM) layer.
8. The computer-implemented method of claim 1 , wherein generating the plurality of embedding vectors representing the set of tokens includes at least one of:
computing an ELMo embedding for each of the set of tokens using a trained ELMo model; or
computing a GloVe embedding for each of the set of tokens using a trained GloVe model.
9. A system comprising:
an online server comprising memory and one or more processors, the memory storing instructions, wherein the instructions, when executed by the one or more processors, cause the one or more processors to perform steps comprising:
receiving an image of a historical document, the image digitized by optical character recognition of the historical document, the historical document comprising a set of non-narrative units, each non-narrative unit describing one or more natural persons that are different from other non-narrative units and each non-narrative unit concerning the same topic as other non-narrative units;
extracting a set of tokens from the historical document, wherein the set of tokens comprising a first token that represents a particular natural person in a particular non-narrative unit, a second token that represents a relationship associated with the particular natural person;
generating a plurality of embedding vectors representing the set of tokens; and
providing the plurality of embedding vectors to a machine learning language model that is trained to divide the set of non-narrative units to individual units; and
a graphical user interface in communication with the online server, the graphical user interface configured to display a location of the particular non-narrative unit in the image.
10. The system of claim 9 , wherein the steps further comprise:
identifying a set of segments in the image of the historical document; and
identifying the set of non-narrative units based on the set of segments.
11. The system of claim 9 , wherein the steps further comprise:
computing a position vector for each of the set of tokens, wherein the position vector indicates a location of a token with respect to a physical reference point within the image.
12. The system of claim 11 , wherein the position vector of a particular token with respect to a particular segment is one of:
at a beginning of the particular segment;
inside the particular segment; or
outside the particular segment.
13. The system of claim 9 , wherein the historical document is a marriage record that lists couple names as the set of non-narrative units.
14. The system of claim 13 , wherein the marriage record includes a plurality of marriage announcements captured from a newspaper.
15. The system of claim 9 , wherein the machine learning language model includes a bi-directional long short-term memory (LSTM) layer.
16. The system of claim 9 , wherein generating the plurality of embedding vectors representing the set of tokens includes at least one of:
computing an ELMo embedding for each of the set of tokens using a trained ELMo model; or
computing a GloVe embedding for each of the set of tokens using a trained GloVe model.
17. A non-transitory computer-readable medium configured to store code comprising instructions, wherein the instructions, when executed by one or more processors, cause the one or more processors to perform steps comprising:
receiving an image of a historical document, the image digitized by optical character recognition of the historical document, the historical document comprising a set of non-narrative units, each non-narrative unit describing one or more natural persons that are different from other non-narrative units and each non-narrative unit concerning the same topic as other non-narrative units;
extracting a set of tokens from the historical document, wherein the set of tokens comprising a first token that represents a particular natural person in a particular non-narrative unit, a second token that represents a relationship associated with the particular natural person;
generating a plurality of embedding vectors representing the set of tokens;
providing the plurality of embedding vectors to a machine learning language model that is trained to divide the set of non-narrative units to individual units; and
displaying a location of the particular non-narrative unit in the image.
18. The non-transitory computer-readable medium of claim 17 , further comprising:
identifying a set of segments in the image of the historical document; and
identifying the set of non-narrative units based on the set of segments.
19. The non-transitory computer-readable medium of claim 17 , further comprising:
computing a position vector for each of the set of tokens, wherein the position vector indicates a location of a token with respect to a physical reference point within the image.
20. The non-transitory computer-readable medium of claim 17 , wherein the historical document is a marriage record that lists couple names as the set of non-narrative units.