User:BushmasterSWA/sandbox

= Speech Analytics (SA) and Business Intelligence (BI) =

Business Intelligence (BI)
Business Intelligence is best defined as a broad category of applications and technologies for gathering, storing, analyzing, and providing access to data to help enterprise users make better business decisions. BI applications include the activities of decision support systems, query and reporting, online analytical processing (OLAP), statistical analysis, forecasting, and data mining.

Types of Data Collection and Analysis
For the purpose of this article, there are two main types of data collection and analysis technologies that can be applied to BI; Statistical Analysis and Data Mining.

Statistical Analysis
Statistical analysis refers to a collection of methods used to process large amounts of data and report overall trends. Statistical analysis is particularly useful when dealing with noisy data. Statistical analysis provides ways to objectively report on how unusual an event is based on historical data. An organization might use statistical analysis on a set of data that has been collected over a period of time to identify trends and correlations. The price fluctuation of a particular product versus region, or time of year, are good examples of this type of analysis.

Data Mining
The data mining activity is one that is seen as one of the most promising applications when dealing with immense amounts of collected data, or big data. Data Mining has been around for quite awhile. In a description from 1996, Simoudis describes Data Mining as "The process of extracting previously unknown, comprehensible and actionable information from large databases and using it to make crucial business decisions." This definition has business flavor and is most useful for business environments. However, data mining is a process that can be applied to any type of data ranging from weather forecasting, electric load prediction, product design, etc.

Data mining also can be defined as the computer-aid process that digs and analyzes enormous sets of data and then extracting the knowledge or information out of it. By its simplest definition, data mining automates the detections of relevant patterns in a database.

One of the features of Data Mining is that it can be applied to any type of data, as long as it is structured. One of these structured data types is Speech, or Voice, such as that resident in phone or video conversations. Many organizations collect or work with Speech data, but do not apply the necessary formatting or structure requirements to make it useful in the decision-making process. This is where Speech Analytics comes into play.

Definition
Speech Analytics is best defined as a method that is used to examine, identify and cross-reference recorded speech to gather useful insight about the speech or the speaker. Language which originated as tool for humans to describe everything in this world has evolved into a complex multi-dimensional medium that is used to express ideas, emotions and much more. If language can be used to describe an emotion, then by analyzing the content of a speech the speaker’s state-of-mind can also be understood.

SA applicability to Business Intelligence
According to DMG Consulting, the compelling reasons for continued use of SA Technology are:
 * It addresses a real and measurable need
 * It delivers quantifiable benefits
 * It is not a replacement for something that came before it
 * There is nothing else like it available in the market
 * It can and is being used in conjunction with other solutions
 * It improves the performance and benefits of other applications

These are all valid statements, but the biggest reason that organizations will continue to adopt SA technology is because technology is constantly changing. We are becoming a society reliant on technology and instantaneous access to data, and this reliance is only going to continue to increase. In past business environments, vast amounts of data was often collected and stored, and was either analyzed right away, analyzed at a later date, or discarded without realizing any benefit to its collection and storage. This will have to change for the systems of the future.

While SA can be of assistance for all organizations utilizing BI systems, they are most useful to call centers or other businesses that are speech-centric. Call centers are centered on speech activity, and therefore can receive the greatest gains from SA implementation. In essence, there is a ton of unstructured data that has been recorded and set aside. SA takes all of that data, structures it into advanced queries, and provides the organization with patterns and information they have never seen before.

SA Technology
The technology behind SA varies in its complexity, but requires the same outcome to address the organizational requirement. Audio data must first be recorded and stored, preferably in a format that can be searched by text or speech driven applications. It is generally agreed upon that there are then two accepted methods of applying SA against recorded calls; Speech-to-text and Phonetic-based.

Speech-to-text
In speech-to-text, the audio words are automatically translated into text for rapid searching by the SA application. While normally faster than phonetic-based SA applications, the information gleaned is only as good as the large vocabulary continuous speech recognition (LVCSR) engine or other technology that matches sounds to text. The false positives (not being what it says it is) are normally very high with this technology, and it would be difficult to base decision-making on the results without human intervention and sampling, thus mitigating the machine-based intentions of the software. This technology is more concerned with ‘word-spotting” than individual elements of speech.

Phonetic-based
In phonetic based search, the speech is broken into a string of phonemes, which are the component parts of a language, and then matched against queries to return any audio files that match the query criteria. This technology seems to get better results in the long run, but is more difficult to establish baseline searches and models. The search engine is only as good as the models it’s using. One of the benefits of this technology is that it can detect subtle changes in the actual audio, like emotion and dialect. These are both discriminators that could be used in advance analytics. .

Pros of using SA Technology
There are many positive aspects to SA integration and utilization, to include:
 * Utilization of data previously not available for analysis – As previously mentioned, many organizations record and store hours of this audio without having the resources or the know-how to search through it. SA provides a mechanism for this critical task.
 * Legal audio discovery – When dealing with audio, it is important that there be a mechanism to search for and retrieve data needed for lawsuit or regulatory purposes.

Cons of using SA Technology
While the use of SA is becoming more widespread, and it is powerful technology for the right organization, it does not come without its faults. Some of the difficulties encountered with SA include:
 * Accuracy of data – The tools involved in SA are only as good as the data it is analyzing. Considerations must be made for compression, recording formats, conversion algorithms, etc.
 * Cost – SA technology normally requires the storage and retrieval capability associated with advance servers and processing systems.
 * Technical infrastructure – It is difficult to introduce this technology into an existing architecture. IT know-how and dedicated support are required to make it work correctly. Is your IT department prepared to run servers and systems overnight to enable the processing power required to analyze and index all of the data?
 * Development know-how – Figuring out how the system works in your environment is a challenge. Who is responsible for the models, for further application development, etc.?
 * Requires commitment / Culture of change – Utilizing SA is a big change for the organization. Leaders must show "buy-in" and acceptance of the new system. Users must not only be trained on how to operate the new applications, but also how it benefits the organization as a whole.

Potential SA Market Growth and Technology Trends
In May of 2012, it was estimated that only 23% of contact centers are deploying SA as part of the BI plan. This number is forecasted to rise to 43% by September of 2013. The number of implementations also continued to grow. For the years 2006-2010, SA implementation grew between 22-106% annually.(6) This is not seen as slowing down. To facilitate this growth, businesses will be looking for advice and best practice information from across the business world. Most of the new innovation will be centered on making the data more accurate. This development will focus on the areas of speech engines, advanced queries, and data aggregation/reporting. There will also have to be improvements in ease of use and organizational implementation. Some of the recommendations to facilitate this new growth and acceptance include:
 * Empower users by providing SA data on their desktops – Much of the current work is server-based, and only gets funneled to the established users in the information chain. As the technology transfers from post-event to real-time, users will have the ability to extract and analyze information on the fly.
 * Tie SA to the identification and reporting of business trends – The current practice of capture and storage of speech data will continue to grow to include business trend information. SA tools will facilitate access to trend data such as customer traffic volume, caller behavior, and complaints and suggestions.
 * True emotion detection – This will assist if determining what strategy to apply to the customer.
 * Contact center performance packages – This is centered on training and understanding of the users. This would offer items such as e-learning, surveys, and quality management and compliance tools.
 * Increased learning abilities of the data detection applications – Many of the SA applications in use today are built on the same algorithms that have existed for many years. New R&D is required to find more creative and accurate ways of performing SA. For instance, Microsoft and Nuance are working on new SA and voice recognition software that is based on a technique called Deep Learning, which is modeled after the human brain.