Computer system and method for plagiarism checking and remediation

1 . A computer-implemented method for identifying and remediating plagiarism, comprising:

causing, by one or more processors, natural language processing (NLP) based segmentation of a text document to obtain a plurality of extracted text segments that are selectively classified as either key segments or non-key segments based on named entity recognition (NER) performed in respect of the extracted text segments;

identifying, by the one or more processors, one or more of the extracted text segments as plagiarized text segments, the identifying comprising: (i) selectively querying only in respect of the extracted text segments classified as key segments, via a first application programming interface (API), a first query platform, and (ii) receiving, in response to the querying, at least one respective candidate matching text segment from the first query platform for at least one of the key segments; and

for each extracted text segment identified as a plagiarized text segment:

generating, by the one or more processors, a set of instructions for a generative artificial intelligence (AI) model that instruct the AI model to rephrase the extracted text segment;

providing, by the one or more processors, the set of instructions to the AI model via a second API; and

receiving a response from the AI model that includes a rephrased text segment generated by the AI model,

wherein, the first query platform includes a vector database comprising vector embeddings of respective reference text segments extracted from literary works, the method further comprising obtaining a respective vector embedding for each of the extracted text segments for which the first query platform is to be queried in respect of;

wherein querying the first query platform in respect of an extracted text segment causes the vector database to be searched to identify reference text segments having vector embeddings that meet a similarity criteria with respect to the respective vector embedding obtained for the extracted text segment, and wherein the respective reference text segment for any reference text vector embedding that meets the similarity criteria is identified as a respective candidate matching text segment for the extracted text segment.

2 . The computer-implemented method of claim 1 wherein identifying the one or more extracted text segments as plagiarized text segments further comprises, by the one or more processors:

(iii) for each extracted text segment for which at least one respective candidate matching text segment is received, identifying the extracted text segment as a plagiarized text segment responsive to determining that a similarity between the extracted text segment and the respective candidate matching text segment meets a similarity threshold.

3 . The computer-implemented method of claim 1 wherein the similarity criteria comprises at least one of: a cosine distance; a Jaccard similarity; a word embedding vector distance; or a Levenshtein distance.

4 . The computer-implemented method of claim 2 , wherein identifying one or more of the extracted text segments as plagiarized text segments further comprises, by the one or more processors:

generating one or more queries via a third API for a second query platform requesting identification of one or more respective candidate matching text segments for the one or more of the extracted text segments; and

receiving a response to the one or more queries.

5 . The computer-implemented method of claim 1 comprising selectively querying in respect of the extracted text segments classified as key segments, a second query platform, and (ii) receiving, in response to the querying, at least one respective candidate matching text segment from the second query platform for at least one of the key segments, wherein the first second query platform comprises a metasearch engine that is configured to provide corresponding queries to and aggregate results from multiple search engines based on the querying.

6 . The computer-implemented method of claim 1 comprising selectively querying in respect of the extracted text segments classified as key segments, a second query platform, and (ii) receiving, in response to the querying, at least one respective candidate matching text segment from the second query platform for at least one of the key segments, wherein the first second query platform comprises a plagiarism check engine.

7 . The computer-implemented method of claim 2 wherein the similarity threshold is based on identifying a defined number of identical consecutive words occurring in the extracted text segment and the respective candidate matching text segment.

8 . The computer-implemented method of claim 1 further comprising, by the one or more processors, for at least one rephrased text segment received from the AI model in respect of one of the extracted text segments:

performing an automated check to identify if the at least one rephrased text segment is a further plagiarized text segment, the performing comprising: querying, via the first API, the first query platform, and determining if a respective candidate matching text segment is received from the first query platform;

when the performing of the automated check to identifies the at least one rephrased text segment as a further plagiarized text segment, generating a further set of instructions for the AI model that instruct the AI model to rephrase the at least one rephrased text segment;

providing the further set of instructions to the AI model via the second API and receiving a further response from the AI model that includes a further rephrased text segment generated by the AI model; and

substituting the further rephrased text segment for the extracted text segment.

9 . The computer-implemented method of claim 1 wherein the AI model comprises a large language model (LLM).

10 . The computer-implemented method of claim 1 further comprises, for each rephrased text segment:

causing, by the one or more processors, a graphical user interface (GUI) to be generated that enables a client device to receive user inputs that accept and amend the rephrased text segment; and

responsive to receiving a user input accepting the rephrased text segment, amending the text document by substituting the rephrased text segment, with any user input amendments, for the extracted text segment.

11 . The computer-implemented method of claim 1 wherein the text document comprises question unit content for an examination question, the question unit content including: a question related to a question scenario; a set of answer choices; for each answer choice, an indication of whether the answer choice is correct or incorrect and an explanation of why the answer choice is correct or incorrect; and identification of at least one reference that supports the set of answer choices.

12 . A computer system comprising one or more processors and one or more memories storing machine-executable instructions thereon which, when executed by the one or more processors, cause the computer system to perform a method for identifying and remediating plagiarism comprising:

causing, by one or more processors, natural language processing (NLP) based segmentation of a text document to obtain a plurality of extracted text segments that are selectively classified as either key segments or non-key segments based on named entity recognition (NER) performed in respect of the extracted text segments;

identifying, by the one or more processors, one or more of the extracted text segments as plagiarized text segments, the identifying comprising: (i) selectively querying only in respect of the extracted text segments classified as key segments, via a first application programming interface (API), a first query platform, and (ii) receiving, in response to the querying, at least one respective candidate matching text segment from the first query platform for at least one of the key segments; and

for each extracted text segment identified as a plagiarized text segment:

generating, by the one or more processors, a set of instructions for a generative artificial intelligence (AI) model that instruct the AI model to rephrase the extracted text segment;

providing, by the one or more processors, the set of instructions to the AI model via a second API; and

receiving a response from the AI model that includes a rephrased text segment generated by the AI model

wherein, the first query platform includes a vector database comprising vector embeddings of respective reference text segments extracted from literary works, the method further comprising obtaining a respective vector embedding for each of the extracted text segments for which the first query platform is to be queried in respect of;

wherein querying the first query platform in respect of an extracted text segment causes the vector database to be searched to identify reference text segments having vector embeddings that meet a similarity criteria with respect to the respective vector embedding obtained for the extracted text segment, and wherein the respective reference text segment for any reference text vector embedding that meets the similarity criteria is identified as a respective candidate matching text segment for the extracted text segment.

13 . The system of claim 12 wherein identifying the one or more extracted text segments as plagiarized text segments further comprises, by the one or more processors:

(iii) for each extracted text segment for which at least one respective candidate matching text segment is received, identifying the extracted text segment as a plagiarized text segment responsive to determining that a similarity between the extracted text segment and the respective candidate matching text segment meets a similarity threshold.

14 . The system of claim 12 , wherein

the similarity criteria comprises at least one of: a cosine distance; a Jaccard similarity; a word embedding vector distance; or a Levenshtein distance.

15 . The system of claim 13 , wherein identifying one or more of the extracted text segments as plagiarized text segments further comprises, by the one or more processors:

generating one or more queries via a third API for a second query platform requesting identification of one or more respective candidate matching text segments for the one or more of the extracted text segments; and

receiving a response to the one or more queries.

16 . The system of claim 12 wherein the first query platform further comprises a plurality of search platforms including one or more of: a metasearch engine that is configured to provide corresponding queries to and aggregate results from multiple search engines based on the queries; and/or a plagiarism check engine.