Professional Documents
Culture Documents
Download textbook Maturity And Innovation In Digital Libraries 20Th International Conference On Asia Pacific Digital Libraries Icadl 2018 Hamilton New Zealand November 19 22 2018 Proceedings Milena Dobreva ebook all chapter pdf
Download textbook Maturity And Innovation In Digital Libraries 20Th International Conference On Asia Pacific Digital Libraries Icadl 2018 Hamilton New Zealand November 19 22 2018 Proceedings Milena Dobreva ebook all chapter pdf
https://textbookfull.com/product/research-and-advanced-
technology-for-digital-libraries-20th-international-conference-
on-theory-and-practice-of-digital-libraries-tpdl-2016-norbert-
fuhr/
https://textbookfull.com/product/research-and-advanced-
technology-for-digital-libraries-19th-international-conference-
on-theory-and-practice-of-digital-libraries-tpdl-2015-poznaa-
sarantos-kapidakis/
123
Lecture Notes in Computer Science 11279
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology Madras, Chennai, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7409
Milena Dobreva Annika Hinze
•
123
Editors
Milena Dobreva Maja Žumer
University College London Qatar University of Ljubljana
Doha, Qatar Ljubljana, Slovenia
Annika Hinze
University of Waikato
Hamilton, New Zealand
LNCS Sublibrary: SL3 – Information Systems and Applications, incl. Internet/Web, and HCI
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface by Program Chairs
This volume contains the papers presented at the 20th International Conference on
Asia-Pacific Digital Libraries (ICADL 2018), held during November 19–22, 2018, in
Hamilton New Zealand. Since its beginnings in Hong Kong in 1998, ICADL has
become one of the premier international conferences for digital library research. The
conference series explores digital libraries as a broad foundation for interaction with
information and information management in the networked information society.
Bringing together the achievements in digital libraries research and development from
Asia and Oceania with work from around the globe, ICADL serves as a unique forum
where the regional drive for exposure of the rich digital collections often requiring
out-of-the box solutions meets global excellence.
ICADL 2018 at the University of Waikato in New Zealand offered a further valuable
opportunity for researchers, educators, and practitioners to share their experiences and
innovative developments. The conference was held at Waikato University in Hamilton,
New Zealand, a city of 140,000 people centered on the Waikato River in the heart of
New Zealand’s rolling pastures. The main theme of ICADL 2018 was “Maturity and
Innovation in Digital Libraries.” We invited high-quality, original research papers as
well as practitioner papers identifying research problems and future directions. Sub-
missions that resonate with the conference’s theme were especially welcome, but all
topics in digital libraries were given equal consideration.
After 20 years, digital libraries have certainly reached maturity. At first glance,
research of digital libraries on the general level is no longer needed and the focus has
moved to more specialized areas – but in fact many new challenges have arisen. The
proliferation of social media, a renewed interest in cultural heritage, digital humanities,
research data management, new privacy legislation, changing user information
behavior are only some of the topics that require a thorough discussion and an inter-
disciplinary approach. ICADL remains an excellent opportunity to bring together
expertise from the region and beyond to explore the challenges and foster innovation.
The 2018 ICADL conference was co-located with the annual meeting of the
Asia-Pacific iSchools Consortium and brought together a diverse group of academic
and professional community members from all parts of the world to exchange their
knowledge, experience, and practices in digital libraries and other related fields.
ICADL 2018 received 77 submissions from over 20 countries, attracting also this
year some submissions from the Middle East, a region that does not have its own
designated digital library forum. Each paper was carefully reviewed by the Program
Committee members. Finally, 20 full papers and six short papers were selected to be
presented, and 11 submissions were invited as work-in-progress papers. The submis-
sions to ICADL 2018 covered a wide spectrum of topics from various areas, including
semantic modeling, social media, Web and news archiving, heritage and localization,
user experience, digital library technology, digital library use cases, research data
management, librarianship, and education.
VI Preface by Program Chairs
Shigeo Sugimoto
Organization
Treasurer
David Nichols University of Waikato, New Zealand
Local Organization
Renee Hoad
Tania Robinson
Bronwyn Webster
Rewarding, But Not for Everyone: Interaction Acts and Perceived Post
Quality on Social Q&A Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Sei-Ching Joanna Sin, Chei Sian Lee, and Xinran Chen
User Experience
1 Introduction
This work has been supported by the European Union’s Horizon 2020 research and
innovation programme under grant 770299 (NewsEye).
c Springer Nature Switzerland AG 2018
M. Dobreva et al. (Eds.): ICADL 2018, LNCS 11279, pp. 3–14, 2018.
https://doi.org/10.1007/978-3-030-04257-8_1
4 S. Mutuvi et al.
the reasons why historical documents still pose a challenge include font variation
across different materials, same words spelled differently, material quality where
some documents can have deformations and unavailability of a lexicon of known
historical spelling variants [1]. These factors reduce the accuracy of recognition
which affects the processing of the documents and, overall, the use of digital
libraries.
Among the NLP tasks that can be performed on digitized data is the extrac-
tion of topics, a process known as topic modeling. Topic modeling has become
a common topic analysis tool for text exploration. The approach attempts to
obtain thematic patterns from large unstructured collections of text by grouping
documents into coherent topics. Among the common topic modeling techniques
are the Latent Dirichlet Allocation (LDA) [3] and the Non-negative Matrix fac-
torization (NMF) [11]. The basic idea of LDA is that the documents are rep-
resented as random mixtures over latent topics, where a topic is characterized
by a distribution over words [30]. The standard implementation of both the
LDA and NMF rely on stochastic elements in their initialization phase which
can potentially lead to instability of the topics generated and the terms that
describe those topics [14]. This phenomena where different runs of the same
algorithm on the same data produce different outcomes manifests itself in two
aspects. First, when examining the top terms representing each topic (i.e. topic
descriptors) over multiple runs, certain terms may appear or disappear com-
pletely between runs. Secondly, instability can be observed when examining the
degree to which documents have been associated with topics across different runs
of the same algorithm on the same corpus. In both cases, such inconsistencies
can potentially affect the performance of topic models. Measuring the stability
and coherence of topics generated over the different runs is critical to ascertain
the model’s performance, as any individual run cannot decisively determine the
underlying topics in a given text corpus [14].
This study examines the effect of noise on unsupervised topic modeling algo-
rithms, through comparison of performance of both the LDA and NMF topic
models in the presence of OCR errors. Using a dataset comprising corpus of
OCRed documents described in Sect. 4, both the topic stability and coherence
scores are obtained and comparison of models’ performance on noisy and the
corrected OCR text is conducted. To the best of our knowledge, no other study
has attempted to evaluate both the stability and coherence of the two models
on noisy OCR text corpora.
The remainder of the paper is structured as follows. In Sect. 2 we discuss
related work on topic modeling on OCR data. In Sect. 3 we describe the met-
rics for evaluating the performance of topic models, namely topic stability and
coherence, before evaluating LDA and NMF topic models in the presence of
noisy OCR in Sect. 4. We discuss the experiment results and conclusion of the
paper with ideas for future work in Sects. 5 and 6, respectively.
Evaluating the Impact of OCR Errors on Topic Modeling 5
2 Related Work
2.1 OCR Errors and Topic Modeling
Topic models aim to discover the underlying semantic structure within large
corpus of documents. Several methods such as probabilistic topic models and
techniques based on matrix factorization have been proposed in the literature.
Much of the prior research on topic modeling has focused on the use of prob-
abilistic methods, where a topic is viewed as a probability distribution over
words, with documents being mixtures of topics [2]. One of the most commonly
used probabilistic algorithms for topic modeling is the Latent Dirichlet Alloca-
tion (LDA) [3]. This is due to its simplicity and capability to uncover hidden
thematic patterns in text with little human supervision. LDA represents top-
ics by word probabilities, where words with highest probabilities in each topic
6 S. Mutuvi et al.
determine the topic. Each latent topic in the LDA model is also represented
as a probabilistic distribution over words and the word distributions of topics
share a common Dirichlet prior. The generative process of LDA is illustrated as
follows [3]:
(i) Choose a multinomial topic distribution θ for the document (according to a
Dirichlet distribution Dir(α) over a fixed set of k topics)?
(ii) Choose a multinomial term distribution ϕ for the topic (according to a
Dirichlet distribution Dir(β) over a fixed set of N terms)
(iii) For each word position
(a) Choose a topic Z according to multinomial topic distribution θ.
(b) Choose a word W according to multinomial term distribution ϕ.
Various studies have applied probabilistic latent semantic analysis (pLSA) model
[4] and LDA model [5] on newspaper corpora to discover topics and trends over
time. Similarly, LDA has been used to find research topic trends on a dataset
comprising abstracts of scientific papers [2]. Both pLSA and LDA models are
probabilistic models that look at each document as a mixture of topics [6]. The
models decompose the document collection into groups of words representing
the main topics. Several topic models were compared, including LDA, correlated
topic model (CTM), and probabilistic latent semantic indexing (pLSI), and it
has been found that LDA generally worked comparably well or better than the
other two at predicting topics that match topics picked by human annotators [7].
MAchine Learning for LanguagE Toolkit (MALLET) [8] was used to test the
effects of noisy optical character recognition (OCR) data using LDA [9]. The
toolkit has also been used to mine topics from the Civil War era newspaper
dispatch [5], and in another study to examine general topics and to identify
emotional moments from Martha Ballards diary [10].
Non-negative Matrix Factorization (NMF) [11] has also been effective in dis-
covering topics in text corpora [12,13]. NMF factors high-dimensional vectors
into a low-dimensionality representation. The goal of NMF is to approximate a
document-term matrix A as the product of two non-negative factors W and H,
each with k dimensions that can be interpreted as k topics. Like LDA, the num-
ber of topics k to generate is chosen beforehand. The values of H and W provide
term weights which can be used to generate topic descriptions and topic mem-
berships for documents respectively. The rows of the factor H can be interpreted
as k topics, defined by non-negative weights for each [14].
|Ri ∩ Rj |
Jac(Ri , Rj ) = (1)
|Ri ∪ Rj |
where Ri denotes the top t ranked terms for the i-th topic (topic descriptor). We
can use Eq. 1 to build a measure of the agreement between two complete topic
models (i.e., Term Stability):
k
1
T S(Mi , Mj ) = Jac(Rix , π(Rix )) (2)
k x=1
where a score of 1 indicates that all pairs of topic descriptors matched together
across the r runs contain the same top t terms [14].
Topic model stability can also be established from document-topic associa-
tions. PNMI determines the extent to which the dominant topic for each doc-
ument varies between multiple runs. The overall level of agreement between a
set of partitions generated by r runs of an algorithm on the same corpus can be
computed as the mean Pairwise Normalized Mutual Information (PNMI) for all
8 S. Mutuvi et al.
pairs:
r
1
PNMI = N M I(Pi , Pj ) (4)
r × (r − 1)
i,j:i=j
In the next section, we use the theory described in this section to determine
the stability and coherence scores of topics generated by LDA and NMF topic
models, from the data described in Sect. 4.1.
4 Experiments
In this section, we seek to apply topic modeling techniques, LDA and NMF, on
the OCR text corpus described below, in an attempt to answer the following
two questions:
Evaluating the Impact of OCR Errors on Topic Modeling 9
(i) To what extent do OCR errors affect the stability of topic models?
(ii) How do the topic models compare in terms of topic coherence, in the presence
of OCR errors?
were computed using Eqs. 3 and 4, respectively. The results of the average topic
stability and the average partition stability are shown in Tables 1 and 2, respec-
tively. Figure 1 provides a graphical representation of LDA and NMF stability
scores on the different dataset categories.
Both models recorded higher average topic stability on the aligned text com-
pared to the raw OCR text. The mean stability on the Gold Standard text was
0.265 and 0.414 while for the noisy OCR text was 0.252 and 0.383 for LDA and
NMF topic models, respectively.
The average partition stability for LDA remained unchanged for both aligned
and raw OCR text. However, NMF recorded a mean partition stability score of
0.117 and 0.114 for the Gold Standard and raw OCR text, respectively.
The results of the average coherence score for LDA and NMF algorithms, on the
noisy and corrected data are presented in Table 3.
The mean coherence score on the aligned OCR text was 0.4737 and 0.3585 for
NMF and LDA algorithms respectively. On the other hand, the mean coherence
using the raw OCR text was marginally lower recording 0.4720 for NMF and
0.3529 for LDA topic model.
5 Discussions
Topic modeling algorithms have been evaluated based on model quality and
stability criteria. The quality and stability of the algorithms was determined
by examining topic coherence and term and document stability respectively.
Overall, the aligned corpus text had higher stability score compared to the noisy
12 S. Mutuvi et al.
OCR input text for both the LDA and NMF topic modeling techniques. The
NMF algorithm yielded the most stable results both at the term and document-
level, as shown in Fig. 1.
On the other hand, the evaluation of topic coherence showed topics from
the corrected corpus were more coherent compared to the original noisy text for
both the LDA and NMF models. As expected, LDA had a lower coherence score
than NMF, which may reflect the tendency of LDA to produce more generic and
less semantically-coherent terms [18]. The difference in average coherence score
between the models was relatively small for the aligned OCR and noisy OCR
text corpus.
6 Conclusions
It is evident from this study that OCR errors can have a negative impact on
topic modeling, therefore affecting the quality of the topics discovered from text
datasets. Overall, this can impede the exploration and exploitation of valuable
historical documents which require use of OCR techniques to enable their digi-
tization. Advanced OCR post correction techniques are required to address the
impact of OCR errors on topic models.
Future research can explore the impact of OCR errors on the accuracy of
other text mining tasks such sentiment analysis, document summarization and
named entity extraction. In addition, multi-modal text mining approaches that
put into consideration textual and visual elements can be explored to determine
their suitability in processing and mining of historical texts. Evaluating the
stability and coherence for different number of topic models can also be examined
further.
References
1. Silfverberg, M., Rueter, J.: Can morphological analyzers improve the quality of
optical character recognition? In: Septentrio Conference Series, vol. 2, pp. 45–56
(2015)
2. Rosen-Zvi, M., Griffiths, T., Steyvers, M., Smyth, P.: The author-topic model for
authors and documents. In: Proceedings of the 20th Conference on Uncertainty in
Artificial Intelligence, pp. 487–494. AUAI Press (2004)
3. Blei, D.M., Ng, A.Y., Jordan, M.I.: Latent dirichlet allocation. J. Mach. Learn.
Res. 3, 993–1022 (2003)
4. Newman, D.J., Block, S.: Probabilistic topic decomposition of an eighteenth-
century American newspaper. J. Assoc. Inf. Sci. Technol. 57(6), 753–767 (2006)
5. Nelson, R.K.: Mining the dispatch (2010)
6. Yang, T.I., Torget, A.J., Mihalcea, R.: Topic modeling on historical newspapers. In:
Proceedings of the 5th ACL-HLT Workshop on Language Technology for Cultural
Heritage, Social Sciences and Humanities, pp. 96–104. Association for Computa-
tional Linguistics (2011)
7. Chang, J., Gerrish, S., Wang, C., Boyd-Graber, J.L., Blei, D.M.: Reading tea
leaves: how humans interpret topic models. In: Advances in Neural Information
Processing Systems, pp. 288–296 (2009)
Evaluating the Impact of OCR Errors on Topic Modeling 13
27. Beitzel, S., Jensen, E.C., Grossman, D.A.: A survey of retrieval strategies for OCR
text collections. In: Proceedings of 2003 Symposium on Document Image Under-
standing Technology (2003)
28. Taghva, K., Nartker, T., Borsack, J., Lumos, S., Condit, A., Young, R.: Evaluating
text categorization in the presence of OCR errors. In: Document Recognition and
Retrieval VIII. International Society for Optics and Photonics, vol. 4307, pp. 68–75
(2000)
29. Agarwal, S., Godbole, S., Punjani, D., Roy, S.: How much noise is too much: a study
in automatic text classification. In: Proceedings of the Seventh IEEE International
Conference on Data Mining, ICDM 2007, pp. 3–12 (2007)
30. Steyvers, M., Griffiths, T.: Probabilistic topic models. In: Handbook of Latent
Semantic Analysis, vol. 427, no. 7, pp. 424–440 (2007)
31. Walker, D., Ringger, E., Seppi, K.: Evaluating supervised topic models in the
presence of OCR errors. In: Document Recognition and Retrieval XX, vol. 8658,
p. 865812. International Society for Optics and Photonics (2013)
Measuring the Semantic World – How to Map
Meaning to High-Dimensional Entity Clusters
in PubMed?
1 Introduction
In digital libraries the increasing information flood requires new and innovative access
paths that go beyond simple term-based searches. This is of particular interest in the
scientific field where the number of publications is growing exponentially [17] and
access to knowledge is getting increasingly difficult, such as to (a) medical entities like
active substances, diseases, or genes and (b) their relations. However, these entities and
their relations play a central role in the exploration and understanding of entity rela-
tionships [2]. Extracting entity relations automatically is therefore of particular interest,
because it bears the potential for new insights and numerous innovative applications in
important medical research areas such as the discovery of new drug-disease associa-
tions (DDAs) needed, e.g., for drug repurposing [14]. Previous work has recognized
this trend and focuses on the recognition of these pharmaceutical entities and their
relationships [3, 6, 9]. What is a DDA? A DDA is in general an effect of drug x on
disease y [3], which means (a) an active substance helps (cures, prevents, alleviates) a
certain disease or (b) an active substance causes/triggers a disease in the sense of a side
effect. Why are DDAs of interest? DDAs are considered as potential candidates for drug
repurposing. Pharmaceutical research attempts to use well-known and well-proven
active substances against other diseases. This generally leads to a lower risk (in the
sense of well-known side effects [4]) and significantly lower costs [4, 6]. Based on the
interests mentioned above, numerous computer-based methods were developed to
derive DDAs from text corpora as well as from specialized databases [9]. The similarity
between active substances and diseases forms the basis here, and numerous popular
methods exists for calculating a similarity between these pharmaceutical entities such
as chemical (sub) structure similarity [8] or network-based similarity [4].
Newer approaches attempt to derive an intrinsic connection between pharmaceutical
entities with a linguistic, context-based similarity [7, 10, 11]. The basic idea is the
distributional hypothesis: a similar linguistic word-context indicates a similar meaning
(or properties) of the entities contained in texts. In this kind of entity-contextualization the
currently popular distributional semantic models (also embedding models or neural
language models) play a major role because they enable an efficient way for learning
semantic word-similarities in huge corpora. However, non-deterministic word-
embedding models like Word2Vec are on one hand popular models for semantic, as
well as analogy tasks, but on the other hand their properties are not fully investigated [15].
In this context, we address the following research questions: (Q1) Is a meaningful
DDA also represented in the word embedding space and can we measure it in terms of
a linguistic distance? (Q2) How can this be measured, evaluated, and what are
meaningful baselines and datasets? (Q3) Since a word-context is learned on the basis of
millions of publications, is new knowledge discovered with this kind of contextual-
ization? (Q4) Are even DDA predictions possible with such models, and if so, how
high is the respective predictive factor?
In this paper we answer all these questions and follow our use case of pharma-
ceutical entities drug/disease and their relations (DDAs). We evaluate the embedded
entities both with manually curated data from specialized pharmaceutical databases as
well as text-mining approaches. In addition, we carry out a retrospective analysis,
which shows that low-distance relations (which previously did not explicitly occur in
documents) will actually occur in future publications with a high probability. This
indicates that a future relation already exists at an early stage in embedded space which
can also help us to reveal a future drug-disease relation. The paper is organized as
follows: Sect. 2 revisits related work accompanied by our extensive investigation of
embedded drug-disease associations in Sect. 3. We close with conclusions in Sect. 4.
2 Related Work
Research in the field of digital libraries has long been concerned with semantically
meaningful similarities for entities and their relations. With a high degree of manual
curation numerous existing systems guarantee a reliable basis for value-adding services
Another random document with
no related content on Scribd:
COMMENTS
CONSTRUCTIVE TREATMENT
Miss Burr was wrong to conclude that it was not her business to
protect the health of her pupils.
There were several beneficial results accompanying the
Chautauqua salute program described above. Those who did good
work were rewarded and those who cheered them were given drill in
the hard task of praising their fellows who had succeeded where they
themselves had failed.
This salute was given in the morning, so that all handkerchiefs
would be clean for it.
Teachers can do no better than tactfully Enlist the
to enlist the aid of mothers in matters of Mothers
cleanliness.
Miss Shaw, who taught the fourth grade in a very poor district in
New York City, organized a mothers’ meeting to convene alternate
Fridays after school hours. At one of these meetings she asked a
trained nurse to talk about cleanliness especially.
After the nurse had clearly explained the danger of infection in the
care of the nose, a mother who was a good shopper was delegated to
take orders for handkerchiefs from all present who needed them for
their children, to buy by the dozen and to deliver to the mothers at
cost.
By teaching and helping the mothers in this way, Miss Shaw
bettered conditions in her own room and established a wholesome
community spirit among her patrons.
CONSTRUCTIVE TREATMENT
COMMENTS
Miss Stow, who taught the fifth grade at Deadwood, made up this
motion song which the children sang occasionally, suiting the proper
motions to each verse:
CONSTRUCTIVE TREATMENT
A wise teacher does not make so much ado about one aspect of
school-room management. Mr. MacFarland could easily have
foreseen that he was introducing another evil along with his reform.
Reverse the process. Arrange that the wastebasket be passed around
mid-way between intermissions according to a definite schedule. Let
the passing of the basket be a privilege that is handed down from one
pupil to another according as they are seated, beginning with the
pupil who sits to the extreme right of the teacher’s desk. One pupil
each day is appointed as guardian of waste. The privilege and honor
will be appreciated, and with a word of caution, the service can be
performed with the very slightest interruption.
The following instructions might be properly given to the children
when the new system is established:
“We are going to save time as much as possible. Four times every
day we shall pass the wastebasket and you may put into it whatever
you have on hand to be thrown away. Nothing is to be said to the
person who carries the basket. He will go quietly, promptly, passing
down the aisles, doing a favor to everyone in the room by his careful
attention to this matter.”
COMMENTS
It is of the utmost importance that the necessary organization of
school-room conduct be established according to principles and
policies that are useful to the child at home and in other situations
outside of school life. He must be taught how to economize his own
time, and the attention of his fellows, how to do necessary things in a
way that shall be both effective and unobtrusive.
While the handling of waste is a small affair for one room in a
school, taken on a larger scale for home, state and nation, the
problem of waste is big enough to command the attention of every
citizen; therefore, to dispose of the matter properly in school is a
valuable lesson in civics and economics.
COMMENTS
COMMENTS
Miss Swainson’s method did not go deep enough. Mere protest will
occasionally effect a change, but not often. If she could have gotten
behind Charlie Owen’s attitude, and found its roots, which were
probably in the attitude of some member of his family or a powerful
older friend, she might have truly converted Charlie to her way of
thinking. His statement that “everybody always laughs at those
Lescinszkies,” might have given her a clue, but she did not realize
that she was only working on the surface.
ILLUSTRATION (RURAL SCHOOL)
—Hughes.
Much school trouble is caused by the purely sportive impulses of
childhood, impulses which are in themselves entirely innocent and
wholesome. One of the most valuable parts of a child’s training is the
acquiring of a set of notions as to appropriateness—the knowledge of
when he may, and when he may not, rightly give rein to his wish to
play. Some children acquire these ideas of propriety readily, and
adapt themselves seemingly without effort to the customs of their
environment; but most children stumble through the period of
adaptation with many backslidings, for the instinct of play is stronger
than the instinct of adaptation to requirement.
But let it be remembered meanwhile that this same play instinct is
one of the strongest allies of the teacher in securing such adaptation,
if the instinct is properly directed. What lesson of politeness,
neatness, unselfishness, protection of the weak, promptness,
responsibility, care of pets, coöperation, chivalry—yes, even of duty
and religion—may not be taught through play! Draining off the play
impulses into these legitimate channels will relieve many a
wearisome, perplexing day for both teacher and pupils, and at the
same time speed on the child toward conscious self-control.
Perhaps the greatest single help in teaching children the voluntary
limitation of their play impulses, is the knowledge that play is only
postponed, not forbidden. Most children have so strong a love of
approbation that they like to do things in the proper way if the
sacrifice be not too great. They are willing to put off their fun, but not
to put it off forever. The teacher who says, “If you’ll wait until recess,
I’ll show you how to play a new game with marbles,” will secure
willing obedience, when she who takes the marbles away has only
sullen submission for her reward. Here the teacher utilizes the
instinctive love of novelty in teaching control of play impulses.
During the period, when conduct is so largely a matter of instinct,
wise teachers play off one instinct against another to the child’s gain,
knowing that some impulses need encouragement and others need to
be inhibited.
(1) “Just mischief.” One of the most First Grade
frequent ways in which the play instinct
expresses itself in the first and even in higher grades is in the little
annoyances which teachers group together under the general term,
mischief. An energetic child, if he is not constantly employed,
naturally vents his energy either in play or in trying to satisfy his
curiosity.
The principle of suggestion alone will often be sufficient to control
the child. The principle of substitution will work equally well.
Coöperation can be correlated with either of these, and expectation
that the child will do what the teacher desires should be used in
whatever method the teacher may adopt.
The mischievous boy will be quiet so long as he is reciting, but
while others are reciting he will immediately hunt for something else
to do. He will drag his shoe on the floor, reach over and touch his
neighbor, pull out a pocketful of string or help himself to his
neighbor’s pencil.
It is not a case for punishment, but a time to apply one or more of
the fundamental principles. The teacher, without even a word or
look, may reach over and draw him to her side, asking him to look on
her book, or better still she may look on his book (suggesting that he
attend to his lesson). She may send him on some little errand—to
bring a book or a piece of crayon—and by the time she whispers,
“Thank you, you did that like a little gentleman,” he will have
forgotten all about his mischief (substitution and approval).
When the class is dismissed, however, the teacher must see that
the mischievous boy is kept busy and his work changed once or twice
within the half hour. She must not fail to show an interest in his
work. The comment, “That is so good, my boy, that I want you to put
it on my desk where we can look at it,” will so elate the child that he
will work industriously to do his best, and doing his best will keep
him busy a long time. When on the playground or in the gymnasium
the teacher must see that he gets plenty of “full-of-fun” play. It will
use up some of his restless energy.
The teacher must have an abundance of busy work ready for the
next restless period that comes, sorting blocks or marbles, straws or
papers, cutting or coloring pictures, putting books on the shelf, and
anything that will keep him innocently busy. Stencil cards, cutting
out words or letters, the distributing of materials for the class to use,
games for the recess and noon periods, frequent story periods during
school hours, interesting lessons, vigorous exercise, generous
approval of every effort to please the teacher—all these will gradually
win the mischievous boy to habits of self-control and industry. All