New technologies driven by Neural Networks provide more and more interesting functions. This time I am going to run a text analysis in such a way, that I will ask questions about the content of a text and the system shall give me (hopefully) correct answers.
I have chosen a text of my preferred author and thinker Dr. Eliyahu Goldratt about "Engines of Disharmony". The text is publicly available on the website of Goldratt Consulting  in Pdf format. The text represents the foreword of the Japanese version of Eli's book "The Choice".
I downloaded the Pdf file and transformed the text into a plain ASCII text. For the analysis, I only used the first four pages of the original document.
Basic text analysis:
- Identified language: English
- Number of words: 1574
- weighted Word Cloud:
- Used letters:
Of course we could add now other funny analytical things, but let us focus on the core of this post:
Find Textual Answer - Neural Network
Mathematica provides a relatively new interesting feature called "FindTextualAnswer". This feature answers questions by quoting the most appropriate excerpts of a text that is presumed to contain the relevant information.
Are you interested on the applied Neural Network? Please, click here.
Let us give it a trial on Goldratt's text "Engines of Disharmony". Beside the answers, I also share the estimated probability P that the answer is correct.
Question 1: "What is the conviction of the author?"
Answer 1: "All my work is based on the conviction that, the underlying concepts and methods used in the hard sciences can, and should, be applied to the social sciences." (P = 0.488071)
Question 2: "What is the opinion of Isaac Newton?"
Answer 2: "One of these underlying concepts, as was first put by Sir Isaac Newton, is that reality is harmonious with itself (a concept which is explained in depth in this book". (P = 0.281488)
Question 3: "What is in the CCPM solution?"
Answer 3.1: "The CCPM solution removes the need for safety at the task level while providing clear priorities and realistic progress information". (P =0.713345)
Answer 3.2: "I started to examine what causes inharmonious relationships between people and why the CCPM solution enables the same environment to enjoy real harmonious relationships". (P =0.556295)
Answer 3.3: "The CCPM application is based on the realization that the deliverer does include enough safety (actually, too much safety) but the safety time is wasted". (P =0.410159)
Answer 3.4: "One of the Theory Of Constraints applications that is rapidly expanding in Japan is the application for project environments, called Critical-Chain-Project-Management or CCPM." (P =0.395317)
Answer 3.5: "Widely spreading teamwork, motivation increasing across the company: I have always wanted our company to be like this! When I developed the CCPM solution I concentrated on trying to decipher the underlying harmony of the project environment itself zooming in on our erroneous assumption that caused the core conflict that destroyed that harmony." (P =0.217434)
Question 4: "Why is it necessary to have harmony?"
Answer 4.1: "It seems to me that the eagerness for harmony is translated into a deep desire and mode of behavior that try to prevent confrontations.
Preventing confrontations is certainly good but not when it is done at the expense of solving the existing conflicts." (P =0.65796)
Answer 4.2: "Rather than looking for the harmony in the situation, in a sense the same harmony one is looking for in art, I started to look on the harmony between people." (P =0.552068)
Answer 4.3: "I'm starting to be, more and more, of the opinion that the eagerness to have harmony, an eagerness that separates Japan from any other nation I know, is paradoxically the number one reason that the desired harmony is not always achieved." (P =0.490903)
Answer 4.4: "I started to examine what causes inharmonious relationships between people and why the CCPM solution enables the same environment to enjoy real harmonious relationships". (P =0.411966)
Question 5: "What is the objective of a project leader?"
Answer 5.1: "It is actually your responsibility that the projects your company commits to are delivered on time". (P =0.487518)
Answer 5.2: "You want to do a good job, to make sure that your project will be on time, within budget and delivering all the promised content" (P =0.449646)
Answer 5.3: "Put yourself in the shoes of a person who is in charge of delivering a task in a project." (P =0.334645)
Answer 5.4: "My impression is that being regarded as a reliable employee is even more important in the Japanese culture" (P =0.330854)
Answer 5.5: "Your only way is to approach their boss, a resource manager, and to convince him to put his people on your project" (P =0.304664)
We can ask much more questions and the system will give us the answers in the same way. But please keep in mind, because the functionality is based on statistical methods, asking the same question in different ways can provide different answers.
Meantime Neural Networks / Deep Learning is extremely powerful. Nevertheless, this new technology has one significant limitation: There is no logic behind!
Moreover, this statement is valid for all kind of neural network algorithms.
Context: "On Monday last week, I bought five bottles of excellent Italian red wine. Unfortunately, I lost 2 bottles during transport from supermarket to my home. A good friend of me endowed me one bottle last weekend.
Question: "How many bottles of red wine do I have now?
Answer: "On Monday last week, I bought five bottles of excellent Italian red wine."(P =0.88973)
I tried to do the same kind of analysis with the text of the business novel "The Goal" written by Eli Goldratt. However this attempt failed. My asssumption is that especially dialogues among peoples are very difficult to interpret by the algorithm. Improvements for specific kinds of text are needed.
What are the benefits of this technology application?
- You can get very easy answers to questions regarding the content of a text.
- The combined content from several text sources can give much more insights about specific topics by one or more authors.
- You can do all this without reading the full text sources.
 Eli Goldratt, "Engines of Disharmony", https://www.goldrattconsulting.com/webfiles/fck/files/Engines-of-Disharmony.pdf, last accessed on July 13, 2019
 Jérôme Louradour, "New in the Wolfram Language: FindTextualAnswer", https://blog.wolfram.com/2018/02/15/new-in-the-wolfram-language-findtextualanswer/, last accessed on July 13, 2019