Natural language, whether spoken or written, has historically been an area of interest in the application of electronic techniques for processing and deduction of information. The term in English is Natural Language Processing and belongs to the field of Artificial Intelligence and Machine Learning. In this article we will limit our discussion to the processing of written open free text. The spoken language is another topic that we can discuss on another occasion.
Imagine the need to read a company's financial results report, or to read a corporate presentation or other report and "understand" the performance of the company, what are the critical points of the report and what summary conclusions can be drawn from this reading. Let us also imagine the financial services industry. Three times a year progress reports are published and at the end an annual report is published by every listed company or by large unlisted companies and organisations.
Put yourself in the shoes of the Stock Exchange Analysts or investment schemes who have to "read" thousands of pages every quarter and "understand" what is going on in these companies. At the same time issue summary reports with their conclusions, the strengths and weaknesses they identify and the trends in economic developments they observe.
Today's intelligent applications are able to extract information from texts with great accuracy, even if they are in free form.
This work until now has been done by eye, by mind and generally has been a manual process. Today it seems that we have the potential to radically change the way we consume information. We can develop intelligent applications that are able to extract information from such texts, even if they are in free form, with great accuracy.
Let's take the example of reading to detect positive, negative and neutral points in a report. The first thing we need is a "dictionary". This dictionary has already been processed by the tools available to analysts, usually in cloud applications, such as Azure. The dictionary has already "tagged" the necessary concepts so that by "reading" the report under review it can identify and isolate "what is going on" in the text. Or even compare the "present" text with previous periods or with other similar reports of similar companies. So it produces one or two pages of report - out of the 200 or 1000 it has read - in a summary and in a way that a human analyst can understand. It has done the heavy work and has saved 99% of the effort and time that a human eye and mind would have put in. It may also have done audit work to identify inconsistencies, errors or discrepancies within the report itself from chapter to chapter or even paragraph to paragraph.
Let's take another example: We write a potential risk as an input to the natural language processing tool, and ask the computer to identify control procedures that would address or mitigate that risk. I.e.: "Computer, tell me (please) what should I do to deal with the possibility of the loss of stock from the warehouse due to obsolescence?" The computer will run through specific areas of control techniques and find the most appropriate ones to address this risk.
NLP tools are not just search engines. They do a full and comprehensive contextual examination on the set of concepts and words contained in a "query" and possible "answers".
Attention: we are not talking here about a search engine investigation but about a full and comprehensive contextual examination of all the concepts and words contained in the "question" and the possible "answers". NLP tools are not just search engines. They read larger sets of words, not neatly arranged in fields and sorted (e.g. not alphabetically but in bulk) and try to "figure out" what the query says and compile processed answers for the user. This is the biggest difference between Natural Language Processing and traditional Data Processing.
And a third example, with the question: "computer, please tell me (please - again, though it's not necessary to plead with the computer!), what kind of feedback do I get about my services on Social Media, and what weaknesses do my clients who post on it identify?" And this is a natural language processing application, which gives as an output detected comments after processing with Machine Learning methods.
Recently, the famous ChatGPT has been introduced to the market, which has made a great impression for its ability to "write" entire essays, motivated by a generic question from a person, who is usually an ordinary everyday user of IT tools. The above mentioned as examples applications can work with older versions of NLP tools. The use of ChatGPT is still under consideration as to HOW it can improve machine learning, in text and natural language processing and understanding.
We at E-On, already have applications like the above that we run on behalf of our customers with excellent results. Our projects aim to save 90% of a person's time in examining text "here and there" as well as producing summary reports from thousands of data in text format. We have at our disposal Risk Dictionaries, Sentiment Algorithms, and other text understanding tools. We are already in the market with NLP solutions with hands-on recommendations aimed at improving Customer Service, addressing climate change issues, risk assessment, financial performance evaluation, and other areas of interest. NLP applications now seem to be coming out to the general public, are affordable and are expected to dominate the understanding and study of business and social issues.
Comments