Speech Recognition and Indexing

Most people can talk faster than they can type. This has an obvious application to normally heavy keyboarding tasks such as writing indexes for books and journals. For most people, typing speeds average around 60 words per minute. However, an average talking speed is typically 150 words per minute.

Besides an increase in input speed, there are other good reasons to become comfortable using Speech Recognition software:

· To reduce repetitive stress injury.

· Users are not limited to sitting in front of a computer—just invest in a wireless mike.

· A digital recorder lets the worker dictate anywhere—take page proofs out on the deck or camping.

Well Reviewed Programs

Two major voice recognition programs that work well with indexing are Dragon Naturally Speaking (Windows) and IBM Via Voice (MacOS). Dragon Naturally Speaking (DNS) version 9.0 runs under Windows XP, and DNS version 9.5 is for Vista. DNS comes in several flavors, Standard, Preferred, Professional, plus versions for specific specialties: Medical and Legal, at increasing price levels and capabilities. The Preferred version is quite adequate for indexers. The balance of this article focuses on DNS, but the general themes are applicable to the Mac version as well.

Accurate recognition of speech requires up-to-date computer equipment. IBM Via Voice on the Mac needs a 1 GHz or better processor. DNS will run well with at least a 1 GHz processor and 1 Gb of RAM. The more speed and RAM in the PC, the better and faster the software will work. A microphone with noise cancelling (blocking out external sounds) is another requirement as is a DNS compatible sound card or an onboard sound system. Compatible equipment can be verified at http://www.nuance.com.

There are two ways of working with DNS while indexing. One method involves the embedded word processor in the program. Using this interface the indexer dictates into DNS, saves the file as a text file, and then imports in their indexing software. This is a faster way to get entries into the file—approximately 50% faster, but not being able to see the full index structure while working probably costs about 15% increase in later editing time.

An alternative is to dictate directly into indexing software. This is more cumbersome because of navigation in the index programs. However, editing is certainly easier since the index structure is constantly visible. Indexers using this speech software can choose one or the other methods based on personal preference and experience.

Current versions of speech recognition software have excellent accuracy rates, approaching 98-99% with proper training. However, the ability to understand the spoken voice is not perfect, but neither is typing. At times, the software will understand a perfectly good word but not what the speaker intended with sometimes humorous results. For example, an indexing term may be dictated as “Older Adults, but the software hears, and enters, “Colder Adults.” Other noted mistakes in user sites are “inscrutable,” which came out “in screw double. and “hyphenate,” rendered as “-8.” At times, a change in the user’s voice such as a cold or sore throat may affect recognition.

When to Use Speech Recognition

Speech recognition and indexing does not function well if the text to be indexed is filled with many non-English names or highly specialized scientific terms. However, dictating index terms works very well with ordinary trade publications, and material in the social sciences, humanities, computer, and technology areas.

As a business expense, the software is relatively inexpensive. The investment should pay off handsomely in increased production of an index, which in turn should decrease turnaround time and increase income.

Talk to your PC

Print Article

Email Article

View All Articles

How to subscribe to feeds