Errata
(i.e. errors in text or diagrams,
not software bugs)




p. 29: simple_coswave.c should be coswave.c. (Thanks to John Fry for pointing this out.)

On page 35 the reader is referred to the file coswave_notes.txt on the CD-ROM, but no such file exists. (Thanks to John Fry for pointing this out.)

p. 37, line 7: When i = 1, cos(i * arg) is NOT 0.157. (Thanks to Burton Rosner for pointing this out.) When i = 1, cos(i * arg) is 0.9563, times 32000 is just over 30601.

In Table 3.2 on page 56, the frequencies listed in the first two columns are half what they should be. E.g., 100 Hz should read 200Hz, and so on. Otherwise they don't match the ratio in the third column. (Thanks to John Fry for pointing this out.)

p. 62, equation 3.4, second line should read B = 2epi w T cos (2 pi f T). (The book has a "/" instead of "f"). (Thanks to Burton Rosner for pointing this out.)

p. 155, figure 5.19. Tibor Kiss (Sprachwissenschaftliches Institut - Ruhr-Universitšt Bochum) has pointed out to me that as well as centre-embedded constructions, this finite state transition network also admits or generates strings such as:

(1)  The malt that the rat that the cat ate that the cat ate ate lay in the house that Jack built.

Prof. Kiss writes: "I am not sure whether you would like to have (1) and other examples included in the language you describe (which should be something like an anbn language with a fixed n). The problem emerges since you can circle e.g. from state 6 to 4 to 5 ... While strings like (1) can be mapped to anc*bn with fixed n, it is somewhat distracting. If you really want the strings in (5-4) to (5-8) and not ones like (1), you need to introduce additional states so that stepping in circles becomes impossible."

p. 168, section 6.3. The file dtw.c was accidentally omitted from the CD-ROM. It can be downloaded from here. (Thanks to Aman Tuladhar for pointing this out to me.)

p. 180. The sum of squared differences in the two tables on this page is 0.431258, not 0.008015. (One I spotted myself!)

p. 192, 16 lines up: "the previous item in the sequence is known as the Markov assumption" (final n in known). (Thanks to David Deterding for pointing this out.)

p. 205, Figure 7.9. The arrow that goes from the node labelled "λ1" to the node labelled "ba" should be labelled "a: P(a | ab)", not "a: P(b | ab)". (Thanks to John Hale for pointing this out.)

p. 246, steps 7 and 8 are (wrongly) the same. (Thanks to David Deterding for pointing this out.) That part of the example derivation should read as follows:
   
Step 7.
Reconsider step 4.
N V DET ADJ* N?  Still not right.
Step 8.
Reconsider step 4 again.
N V ADJ* N?  Still not right.
Step 9.
etc.