Evaluating theses using tree structures
An example system includes a processor to receive a plurality of object aspects of an object to be evaluated using a process, a structure of the process, a plurality of extracted facts from documents, a tree related to the plurality of object aspects and the structure, and a thesis for each leaf in the tree. The processor is also to relate the extracted facts to the theses in the tree. The processor is to generate a score for each leaf corresponding to a fact in the tree. The processor is to generate a thesis score and a thesis summary for each thesis based on the scores and the summaries of related facts for each thesis. The processor is to further generate a final score for the object based on the thesis scores.
1. A system, comprising a processor to:
receive a plurality of object aspects of an object, a root thesis comprising a prediction about the object to be evaluated, a tree related to the plurality of object aspects, and a thesis for each leaf in the tree, wherein each thesis is related to the root thesis;
receive a plurality of extracted facts from a plurality of documents in response to receiving the plurality of documents from an information retrieval process, split each document into sentences, and compute a relatedness score between each sentence and each thesis, wherein sentences comprising a relatedness score above a threshold score are to be extracted as the extracted facts for each thesis;
relate the extracted facts to the theses in the tree, wherein each internal node in the tree represents an aspect with internal nodes representing higher-level aspects connected as parents to internal nodes representing lower-level aspects, wherein the internal nodes representing lowest-level aspects are parents of theses, and wherein the nodes representing theses are parents to the leaves representing the facts;
generate a score and a summary for each leaf corresponding to a fact in the tree;
generate a thesis score and a thesis summary for each thesis based on the scores and the summaries of related facts for each thesis; and generate a final score for the root thesis about the object based on the thesis scores.
2. The system of claim 1 , wherein the theses comprise a base thesis corresponding to a leaf in the tree and a complex thesis corresponding to an internal node in the tree.
3. The system of claim 1 , wherein the thesis score for at least one of the theses is to be generated based on a thesis score of at least one child thesis connected to the thesis.
4. The system of claim 1 , wherein the processor is to train a classification model comprising a recursive neural network trained to process dynamic theses generated per object.
5. The system of claim 1 , wherein the tree comprises a plurality of nodes representing the object aspects and one or more leaves connected to each node, wherein the leaves represent facts.
6. The system of claim 1 , wherein the process comprises a due diligence process and the object comprises a company.
7. A computer-implemented method, comprising:
receiving, via a processor, a plurality of object aspects of an object, a root thesis comprising a prediction about the object to be evaluated using a process, a structure of the process, a tree related to the plurality of object aspects and the structure, and a thesis for each leaf in the tree, wherein each thesis is related to the root thesis;
receive a plurality of extracted facts from a plurality of documents in response to receiving the plurality of documents from an information retrieval process, split each document into sentences, and compute a relatedness score between each sentence and each thesis, wherein sentences comprising a relatedness score above a threshold score are to be extracted as the extracted facts for each thesis;
relating, via the processor, the extracted facts to the theses in the tree, wherein each internal node in the tree represents an aspect with internal nodes representing higher-level aspects connected as parents to internal nodes representing lower-level aspects, wherein the internal nodes representing lowest-level aspects are parents of theses, and wherein the nodes representing theses are parents to the leaves representing the facts;
generating, via the processor, a score and a summary for each leaf corresponding to a fact in the tree;
generating, via the processor, a thesis score and a thesis summary for each thesis based on the scores and the summaries of related facts for each thesis;
generating, via the processor, a final score for the object based on the thesis scores.
8. The computer-implemented method of claim 7 , wherein generating the thesis score and the thesis summary is performed recursively beginning at leaves of the tree.
9. The computer-implemented method of claim 7 , wherein generating the thesis score and the thesis summary is based on a thesis score and a thesis summary of a child thesis.
10. The computer-implemented method of claim 7 , wherein generating the thesis score comprises generating a vector representation.
11. The computer-implemented method of claim 7 , wherein generating the thesis score comprises generating a scalar score.
12. The computer-implemented method of claim 7 , wherein generating the final score comprises calculating a weight for a thesis based on a dot product of an attention vector and a thesis representation.
13. A computer program product for analyzing theses, the computer program product comprising a computer-readable storage medium having program code embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program code executable by a processor to cause the processor to:
receive a plurality of object aspects of an object, a root thesis comprising a prediction about the object to be evaluated using a process, a structure of the process, and a plurality of extracted facts from documents, a tree related to the plurality of object aspects and the structure, and a thesis for each leaf in the tree;
receive a plurality of extracted facts from a plurality of documents in response to receiving the plurality of documents from an information retrieval process, split each document into sentences, and compute a relatedness score between each sentence and each thesis, wherein sentences comprising a relatedness score above a threshold score are to be extracted as the extracted facts for each thesis;
relate the extracted facts to the theses in the tree, wherein each internal node in the tree represents an aspect with internal nodes representing higher-level aspects connected as parents to internal nodes representing lower-level aspects, wherein the internal nodes representing lowest-level aspects are parents of theses, and wherein the nodes representing theses are parents to the leaves representing the facts;
generate a score and a summary for each leaf corresponding to a fact in the tree;
generate a thesis score and a thesis summary for each thesis based on the scores and the summaries of the related facts for each thesis; and
generate a final score for the object based on the thesis scores.
14. The computer program product of claim 13 , comprising program code executable by the processor to recursively generate thesis scores based on child thesis scores generated for child theses of each thesis.
15. The computer program product of claim 13 , comprising program code executable by the processor to calculate a weight for a thesis based on a dot product of an attention vector and a thesis representation.
16. The computer program product of claim 13 , comprising program code executable by the processor to train a neural network to generate confidence scores for each of the theses.
17. The computer program product of claim 13 , comprising program code executable by the processor to generate thesis scores and thesis summaries for each level of the tree beginning with a lower thesis level.