File Converter processes various types of files (PDF, DOCX, TXT) and extracts categorized pieces of text. This helps unify documents into a single format and prepares them for ingestion. This is often done to facilitate the use of a particular NLP tool or software, or to allow the file to be used on a different platform or system.

Document Ingestion is typically used in conjunction with the File Converter module. The text is absorbed by the Soffos AI and prepared for use by the rest of the modules. Document ingestion is an important step in NLP as it allows the data to be cleaned and prepared for further analysis, such as text classification or emotion detection. This is often the first module used in any NLP project, and it is crucial to have a well-defined and reliable ingestion process to ensure that the data is of high quality and ready for further analysis.

The Emotion Detection module requires a body of text, a selection of emotions to detect and instruction on how to segment the text. It then matches each segment to one of the assigned emotions, (e.g., frustration, excitement, worry neutral, etc.). This module can help track customer satisfaction, and it is more insightful than simple sentiment classification.

This module will return any snippets of content that are offensive or contain profanity in a body of text. It can be useful in aiding content filtration, moderating online content, protecting users from harmful content, and ensuring compliance with ethical and legal standards. Additionally, it can help moderators on websites or other platforms decide which words to censor. It can also help flag inappropriate user comments.

The Logical Error Detection module identifies errors in the logical structure of a natural language statement or text which can occur due to incorrect grammar, incorrect word choice, or incorrect syntax, as well as consistencies in given assertions. The module also explains these errors in order to improve the clarity and accuracy of the text. It can help scan dense bodies of text to check for logical fallacies.

Also referred to as Entity Detection, this module identifies and labels entities in a given body of text. These can include the names of people, places, things, etc. This module can help users swiftly understand the subject or theme of a body of text and group text based on relevancy. Example application uses include human resource and customer support operations among other things.

The Natural Language Search module allows users to sift through thousands of ingested documents by asking questions. The module returns relevant passages that answer these queries and returns their origin in the documents. The entire process takes a matter of seconds. This module allows users to extract data from a document using natural language instead of searching for keywords.

This module is useful for testing user knowledge retention and recall capacity. It works by extracting declarative statements from a body of text and rephrasing them into questions. The answers to these questions are the original statements. This can help users study content knowledge better going beyond rote memorization and providing comparative review of free form answers to generated answers to determine accuracy.

This module works by simplifying complex sentences and difficult words in a given body of text. It does not lose information or alter the meaning of the original text in the process but makes it easier for readers to understand. The level of complexity can be altered based on user preference. This module can help users comprehend dense text by simplifying it.

The Question Answering module requires a question and a body of text to query. Not only does it return brief and natural responses to complex questions, but it also identifies ambiguities and offensive statements and responds appropriately to random inputs like greetings. This module can help users find suitable answers from a customized base of information.

The summarisation module condenses a body of text to a given length (number of sentences) while retaining the original meaning. It can blend facts together and produce an original, natural language summary. This can help condense information for users and create more coherent research paper summaries.

This module is useful for determining keywords and subjects in a body of text. It works by analysing the body of text and suggesting tags for it. This process is usually performed to extract specific information from the text or to perform further analysis on the text. Tag generation is useful in modern SEO and conversion tracking for online advertising tools.

STT systems are prone to making mistakes when the speaker has a heavy accent or is speaking unclearly. The Transcript Correction module finds errors made by STT systems when converting voice to words and corrects them to what the user most likely intended to say. This module can be useful for live captions for video conferencing software, or to improve the performance of various STT services.

This module returns snippets from a body of text that contains ambiguities. It detects many types of ambiguities, including those involving word tense and pronoun issues. Additionally, it returns questions to aid the user in resolving the ambiguities. This module can be useful in evaluating a user’s linguistic coherence.

This module returns any parts of a text that contain contradictions. It can also optionally return questions that the user should ask to resolve the contradictions. This can be useful in assessing a user’s grasp on the subject matter of a given body of knowledge. Soffos AI’s Contradiction Detection module uses a hybrid technique, combining both rule-based and machine learning approaches to detect contradictions.