Chatbots

Have you ever used a website and clicked a little live chatbox that popped up on the side of the screen? Or maybe you have recently asked a smart speaker what the weather is.

If so, you have used a chatbot. A chatbot is a software designed to simulate human speech. Chatbots have many benefits, but there are limitations to their power, so it is important to understand how websites and apps use them.

History of Chatbots

In the mid-1960s, a professor at the Massachusetts Institute of Technology named Joseph Weizenbaum created a computer program that mimicked human conversation. This program was called ELIZA and is considered the first chatbot. To create this chatbot, Weizenbaum used pattern matching, a computer science process involving checking sequences of data for patterns and then matching those patterns.

Fig. 1 - The first chatbot was called ELIZA.

The goal of ELIZA was to trick users into believing that they were conversing with a real person. It also had a therapeutic approach to the conversation, asking open-ended questions and using pattern matching to respond with relevant follow-up questions. More complex chatbots were not developed until around 2009 when a chatbot called WeChat launched in China. Many social media platforms like Facebook launched chatbots in the years following. Today, there are many forms of chatbots, including ones that engage in written conversations with users and even speakers that communicate verbally.

Types of Chatbots

Most chatbots are either rule-based chatbots or Natural Language Processing (NLP) chatbots.

Rule-Based Chatbots

Rule-based chatbots can respond to simple commands. They follow a set of pre-set rules that are established when they are programmed. Since they cannot adapt to conversations, they typically involve response buttons like “yes” and “no,” and it is not possible to have an open-ended conversation with them.

Natural Language Processing Chatbots

Chatbots that are able to have an open-ended conversation with users are typically Natural Language Processing (NLP), a branch of artificial intelligence (AI). NLP is the process of computer programs interpreting human language. NLP chatbots are able to interpret and respond to unstructured human language using the following steps:

Step 1: Lexical Analysis

First, the program identifies the parts of paragraphs, sentences, phrases, and words. For instance, it would identify adjectives and nouns in a sentence.

Step 2: Syntactical Analysis

Next, the program analyzes syntax or word order. This includes identifying what part of the sentence is the subject and what part is the predicate.

Step 3: Semantic Analysis

Then, the program has to consider semantics, the literal definition of words. This is tricky because many words have multiple meanings. The program thus has to use the syntactical structure identified in the previous step to determine the definition of each word the user uses.

Fig. 2 - There are five main steps to NLP.

Step 4: Discourse Integration

The program also has to consider how each sentence typed into the chatbot fits into the ones before it, or in other words, how it integrates with the discourse.

Step 5: Pragmatic Analysis

Even with complex analytical abilities, programs still need to understand words in their proper context, a concept called pragmatics. For example, if someone asks a chatbot to lend a hand with something, the chatbot needs to understand that the person meant this metaphorically and is asking for help. Although pragmatic analysis is a step in NLP, chatbots still need to pick up on specific usages of words like idioms and tones like sarcasm.

Benefits of Chatbots

Chatbots have many benefits, mainly related to productivity. These benefits include the following:

Rapid Responses

Unlike humans, chatbots can work twenty-four hours a day. This means that chatbots allow companies to respond to customers or users to reach out to them anytime and receive some help.

Cost

Chatbots can perform many basic customer service functions, saving companies money they would otherwise have to spend on paying employees.

Research

Chatbots help streamline large-scale data collections in research projects. Instead of humans having to go and collect and analyze huge amounts of data, chatbots can ask questions in both qualitative and quantitative research studies.

Fig. 3 - Chatbots can save companies time and money.

Risks of Chatbots

Chatbots have a few risks, both in terms of security and company. Using chatbots can make companies susceptible to hackers, and if the program is not properly encrypted, this could put users’ personal data at risk. In addition, relying on chatbots to handle communication with customers can lead to companies losing consumer support.

As research shows, people prefer communicating with humans, so they might get frustrated that a chatbot does not understand them and prompt them to switch to another company or stop using a company or product altogether. To prevent such actions, creators of chatbots should employ quality assurance tools to continually improve their system, avoid offensive messages, and ensure effective communication.²

Importance of Chatbots

Understanding how chatbots interpret human language can help people understand what words and phrases to use to ensure effective communication. For instance, if a chatbot user knows that it is difficult for chatbots to interpret idioms and metaphors, they can select simpler language to get their point across.

¹ CGS. “Customer Service Chatbot & Channels Survey.” 2019.

² Eleni Adamopoulou and Lefteris Moussiades. “Chatbots: History, Technology, and Applications.” Machine Learning with Applications. 2020.