Results and Analysis

The Language Recognition Program was evaluated by running scans of 200 text excerpts. The current version of the program supports four languages: English, French, Spanish, and German. Fifty text excerpts were chosen for each language from various online sources such as Project Gutenberg and Wikipedia. The results of the scans are then compared to the actual languages that the excerpts are written in. Accuracy ratings were calculated from these results. The results of the evaluation phase are shown below in a table.

A wide variety of text excerpts were chosen. The excerpts were about various topics and this was used to see if the program could handle differing situations. Some of the topics included: everyday objects, biographies, literature, and scientific descriptions. These text excerpts were approximately a page long. Note that other text excerpts were used for developmental testing.

The results are excellent. For three of the four languages the accuracy rating was 100%, and for Spanish the accuracy rating was 98%. This shows that the Language Recognition Program and its algorithms are very effective at determining the language of a text document. Furthermore, it handled all types of topics well.

It was also observed that the program was very fast in processing the text excerpts. The Language Recognition Program was scanning at the rate of 50 000 words per minute (on a computer with a 2.0 GHz processor). This was determined by timing 10 scans of long text excerpts and taking the average of these times. These text excerpts were at least 100 000 characters long. In summary, the Language Recognition Program was able to determine the language of a text excerpt accurately and quickly.