Natural Language Processing (NLP)

What is Natural Language Processing?

Understanding human language is a complex subject, since we express ourselves in various ways both orally and in writing. There are hundreds of languages, each with its own syntax, grammar, slang, etc. Furthermore, when writing, we tend to use abbreviating and drop punctuation. Other challenges we might face when analyzing speech include regional accents, mumbling, and terms acquired from other languages.

Natural Language Processing is a branch of Artificial Intelligence that assists computers in understanding, analyzing and interpreting human Language. This branch allows computers to communicate with humans in their own language and assess unstructured data related to language. For example, NLP enables computers to read text, hear speech, interpret it, assess emotion and define which parts of the text is more important.


What types of data can be analyzed and reviewed?

Structured data- data organized in orderly format. For example, data in SQL tables with columns and rows. They feature a relation index and can be easily charted.

Semi-structured data- data which isn't stored in a structured database yet features some organizational properties that simplify the analysis. A certain processing can arrange them in a certain context. For example, XML data.

Unstructured data- data which is not organized in a predefined manner or is not stored in a predefined database, from texts excerpted from medical records to texts published on social media channels.

Nowadays, computers analyze more language-based data than humans, considering the vast amount of unstructured data generated on a daily basis.

Automation is critical for text and speech data analysis.


What is NLP used for?

NLP technology plays a great role in Artificial intelligence applications, some of which implement machine learning combined with NLP for risk analysis, customer behavior review, Chatbots and fraud detection.

One of the most common uses of NLP is aimed at the various search apps, both audio and textual (for further details, click here). The search engine uses NLP to analyze the query and retrieve the most relevant answers. Users used to have to search by keywords. Nowadays, you can simply state complex sentences that then serve as the basis for your search. When searching by voice, the search questions are said, then recorded and processed by an interface which converts speech into text while filtering background noises, then analyzed for the different phonemes, adapted to words then finally converted to a sentence which becomes query. This query is sent to the system, which in turn sends the appropriate answer. In an age in which people can speak to Siri and Alexa as freely as they speak to one another, many people can benefit greatly from these apps.

These abilities can be used for other applications, too. For example, Facebook uses a tool titled DeepText to analyze posts published on the social network. DeepText is a text analysis engine that can detect positive and negative emotions expressed in written posts, automatically tag and retrieve themes from texts. The objective is to reach insights/conclusions from a text and in some cases even convert it to actual action. An example of this implementation is Chatbots response to a spoken question).


Google has its own NLP tools, including:


API Natural Language Cloud- this tool enables exposing the structure and meaning of the text. It allows retrieving information on people, locations and events to understand the information transferred through social media and client conversations. This tool allows classifying files by 700 predefined categories.

AutoML Natural Language- a machine learning NLP-oriented tool which enables creating personally adapted models for organizations to categorize, retrieve, and detect different elements in a text.

Here’s how it works:

  1. The user loads the various data into the computer (the more varied the examples, the more accurate the process)
  2. Runs the module adapted to their needs
  3. Assesses the received results. In case the received insights are insufficiently precise, update the module, then review the results again.

Source: https://cloud.google.com/natural-language/#how-automl-natural-language-works 


NLP and Knowledge Management

Both AI tools (such as Natural Language Processing) and Knowledge Management tools center around treating knowledge and information as a central component. Both are essentially intertwined. Without reliable knowledge and initial charting of the required templates, the machines would not be able to elaborate, create or implement knowledge optimally. Furthermore, the field of KM has generated the strategic processes that allow organizations to make the knowledge created with their automatic tools optimally accessible to all relevant parties.





Natural User Experience

A natural user interface, as opposed to a man-computer user interface, is based on the natural ways people communicate with each other and their environment and refers to body gestures, movement, facial expressions and touch in order to create an experience of full control of the technology.

Thanks to the late Steve Jobs, that successfully instilled within us an addiction for control and navigating with our fingers, an entire industry of user-friendly, lightweight touch screens allow control of interfaces using our bodies. The main advantage of a natural user interface is that it is appropriate for an extremely wide range of users. It is amazing to see how natural using touch screen is for users from one to one hundred.


The mapping process is a stage which consists of reviewing and prioritizing knowledge needs. During this process, we encounter key factors in the organization or unit and inquire their business "pains" that can be improved if knowledge is managed:

What information and knowledge do they lack in order to perform their job? What information and knowledge do they hold and find important to share with others? During need-mapping interviews, we try to locate challenges and strengths that might contribute to the success of a Knowledge Management activity, inhibit its process or even cause its failure in order to understand how much will the solutions we wish to reach actually benefit the organization and fulfill its business objectives. We will also try to understand how much are these solutions applicable culturally and what are their chances of success. The product at this point should be an indication towards different solutions while presenting a prioritization for their implementation. The prioritization is critical since we always have more knowledge and knowledge needs than the organization can manage, certainly at once.


Communication in the virtual realm is mainly textual. Therefore, the benefits of facial expression, tone and intonation are lost to us. The direct result can be poor communication which can lead to misunderstandings and, in turn, some hurt feelings.

This anonymity can be perceived as a lack of commitment to one another. Nevertheless, most internet users expect that agreements, laws and conventions that exist in the 'real world' will apply in the virtual realm as well. This realm should be seen as extension of the physical reality rather than a new, separate world with a separate set of rules.

This understanding has led to the newly coined term Netiquette, an amalgam of 'net' and etiquette. This term refers to standards and rules of conduct to be followed in this virtual situation.

Its purpose is defining a behavioral standard for online conduct by determining what is permitted, what is prohibited and what is fair. Its ultimate goal is to enable safe use of the web and positive, accurate communication with other users. This universal set of rules is relevant even when national borders are seemingly nonexistent. It addresses the upholding of individual rights, copyright infringement, harassment on the web, and the prevention of child exploitation, etc.




The term Newsgroup describes a database of messages sent by users who are located in different geographical locations, yet share a certain field of interest. The Newsgroup can be compared to an email list, except for the fact that its messages are transmitted through an internet application (and not a mailing server), are exposed to the public and the individual does not need to appear in any mailing list in order to participate in the discussion.

A Newsreader is a type of software intended for organizing the messages received through the newsgroup. The software enables to sign up to a newsgroup and send messages to the group and receive messages from it. In this manner, the tool enables to create a 'newspaper' personally adapted for the individual and thus become a 'subscriber' and remain updated on news fields of interest.

For example, if there is a new newsgroup which interests the user, he must update the newsreader software and subscribe to this team. The newsreader software makes the team accessible to him/her and organizes the messages for him/her. Among other things, it stores information regarding messages which have already been read.


A critical activity when setting up Knowledge Management solutions is marketing the solution. Many of us tend to think that marketing is an activity which is completed when the solution is launched, yet in reality marketing the solution is an ongoing activity which should be performed regularly. One of the most successful marketing tools, especially when dealing with portals, knowledge websites and communities is the newsletter. An electronic newsletter is a structured and designed mail message distributed to all potential users. It is used to acknowledge users as to new central subjects presented in the website/portal/community, those which we especially wish to communicate.

Using the newsletter applies a vital principle of pushing information to users; information which the users are sometimes unaware exists since they are not exposed to it.


Hereby are some principles that will assist us in intelligently using a newsletter:



Nowadays, due to the increase of documented knowledge and information, there is a rivalry between the Push and Pull mechanisms.

Using Pull mechanisms implies that the worker reaches the knowledge through his/her own initiation. Its main disadvantage is that the user doesn't always know what he/she doesn't know and therefore doesn't know what to request.

Using Push mechanisms implies that the knowledge reaches the worker, pushed to him/her by the organization. Its main disadvantage is its violence. It is barging and annoying, is insufficiently focused and therefore includes lots of "junk".

The notification is a good combination of elements from both. In an assigned area, messages are accumulated and organized according to subjects (in order to substantially reduce their amount) about knowledge and information the user should know. A little Push, but moderate. If the user is interested, he/she can perform the pull himself/herself for subjects indeed relevant.