This paper brie y describes the systems presented by the Software Technologies Working Group (http://gtts.ehu.es, GTTS) of the University of the Basque Country (UPV/EHU) to the Spoken Web Search (SWS) task at MediaEval 2013. GTTS systems consist of four main modules: (1) feature extraction; (2) speech activity detection; (3) DTW-based query matching; and (4) score calibration and fusion. The most remarkable contributions are the use of phone loglikelihood ratio features, the normalization of the DTW distance matrix and the calibration/fusion approach (which is imported from language/speaker veri cation).
The MediaEval 2013 Spoken Web Search (SWS) task consists of searching for a spoken query within a set of audio documents [4]. The locations and durations of all the occurrences of spoken queries in the audio documents must be obtained. System performance is primarily measured in terms of the Average Term-Weighted Value (ATWV) [5], but also in terms of a normalized cross-entropy metric and the processing resources (real-time factor and peak memory usage) required by the submitted systems [6]. For more details on the SWS task at MediaEval 2013, see [2]. 2.1
The Brno University of Technology (BUT) phone decoders for Czech, Hungarian and Russian [7] are applied to decode both the spoken queries and the audio documents. BUT decoders are trained on 8 kHz SpeechDat(E) databases recorded over xed telephone networks, containing 12, 10 and 18 hours of speech and featuring 45, 61 and 52 units for Czech, Hungarian and Russian, respectively (three of them being non-phonetic units that stand for short pauses and noises).
Given an input signal of length T , the decoder outputs
the posterior probability of each state s (1 s S) of each
unit i (1 i M ) at each frame t (1 t T ), pi;s(t),
where M is the number of units and S the number of states
per unit. The posterior probability of each unit i at each
frame t are computed by adding the posteriors of its states:
pi(t) = X pi;s(t)
8s
(
Given an audio signal, Speech Activity Detection (SAD) is performed by discarding those phone posterior feature vectors for which the non-speech posterior is the highest. The remaining vectors, along with their corresponding time o sets, are stored for further use, but the component corresponding to the non-speech unit is deleted. If the number of speech vectors is too low (in this evaluation, that threshold was arbitrarily set to 10, that is, 0.1 seconds), the whole signal is discarded, to save time and to avoid false alarms. 2.3
Given two SAD- ltered sequences of feature vectors corresponding to a spoken query q and a spoken document x, the cosine distance is computed between each pair of vectors, q[i] and x[j] as follows: d(q[i]; x[j]) = log with dmin(j) = min d(q[i]; x[j]) and dmax(j) = max d(q[i]; x[j]).
i i In this way, matrix values are all comprised between 0 and 1, so that a perfect match would produce a quasi-diagonal sequence of zeroes.
The best match of a query q of length m in a spoken document x of length n is de ned as that minimizing the average distance in a crossing path of the matrix dnorm. A crossing path starts at any given frame of x, k1 2 [1; n], then traverses a region of x which is optimally aligned to q (involving L vector alignments), and ends at frame k2 2 [k1; n]. The average distance in this crossing path is:
L
davg(q; x) = 1 X dnorm(q[il]; x[jl]) (
L l=1
where il and jl are the indices of the vectors of q and x
in the alignment l, for l = 1; 2; : : : ; L. Note that i1 = 1,
iL = m, j1 = k1 and jL = k2. The minimization operation
(
This procedure is iteratively applied to nd not only the
best match but also less likely matches in the same
document. To that end, a queue of search intervals is de ned
and initialized with (1; n). Let us consider an interval (a; b),
and assume that the best match is found at (a0; b0), then
the intervals (a; a0) and (b0; b) are added to the queue (for
further processing) if: (
Under the extended (multiple examples) condition, only the examples passing SAD ltering (i.e. with enough speech samples) are considered for each query. The longest example is taken as reference and DTW-aligned to the other available examples. Finally, the vectors aligned at each frame are averaged and a single average example is obtained and processed as in the required (single example) condition.
System scores are transformed according to [1], which is an adaptation of the discriminative calibration/fusion approach commonly applied in speaker and language recognition.
First, the so-called q-norm (query normalization) is applied, so that zero-mean and unit-variance scores are obtained per query. Then, if n di erent systems are fused, detections are aligned so that only those supported by n=2 or more systems are retained for further processing (this is known as majority voting validation). Let us consider one of such validated detections, corresponding to a query q; if a system A does not provide a score for it, we use instead the minimum score that A has output for q. The same value is assigned to missed detections and non-target trials. In this way, a complete set of scores is prepared, which besides the ground truth (target/non-target labels) can be used to discriminatively estimate a linear transformation that produces well-calibrated scores that can be linearly combined to get fused scores. Under this approach, the Bayes optimal threshold |given by the e ective prior (0:0148 for this evaluation)| is applied. The BOSARIS toolkit [3] is used to estimate and apply the calibration/fusion models.